arXiv:hep-th/9908171v1 26 Aug 1999 orsodne oegnrlapcsaedsusd hnt Then in discussed. instantons i are between chapter aspects relation final the general The Some detail. func in correspondence. Green reviewed sixteen-point are multiplet and a current eight- semiclassical four-, the of in contr calculations operators one-instanton instant composite the to invariant of introduction gauge computation an the after theories, chapter, gauge fourth the relate In subtleties and out. approach compute superfield th are the the fields and/or in elementary lation presented of functions is Green theory four-point the of properties perturbative n tteqatmlvl h eodcatri eiwo the of review a is chapter second The level. N quantum fou the in at theories supersymmetric and rigid of properties general the of aspects various reviews dissertation This optdi yeIBsrn hoyi efre n agreemen shown. and is performed effects is non-perturbative theory of gener string D-instanton IIB and type theory in field computed boundary to the contributions in instanton operators between comparison The theory. atclri eainwt h d/F correspondence. AdS/CFT the with relation in particular N 4SMter ne consideration. under theory SYM =4 rgnlrslsaerpre ncatr ,4ad5 system A 5. and 4 3, chapters in reported are results Original h rttocatr r nrdcoy h rtoecontain one first The introductory. are chapters two first The 4sprymti agMlstheory Yang–Mills supersymmetric =4 n h d/CTcorrespondence AdS/SCFT the and iatmnod iia nvri` iRm TrVergata” “Tor Roma Universit`a di Fisica, di Dipartimento ....-Szoed oa“o Vergata” “Tor Roma di Sezione - I.N.F.N. i el iec cetfia 1 Scientifica, Ricerca della Via 07 oa ITALY Roma, 00173 tfn Kovacs Stefano N 4SMadDisatneet ntp I string IIB type in effects D-instanton and SYM =4 Abstract N 4sprymti agMlster in theory Yang–Mills supersymmetric =4 iesosbt tteclassical the at both dimensions r oteguefiigaepointed are gauge-fixing the to d in foeaosi h super- the in operators of tions poiaini eotd The reported. is pproximation btost re ucin of functions Green to ibutions supersymmetric in calculus on sn h opnn formu- component the using d r hpe.To,tre and three- Two-, chapter. ird re ucin fcomposite of functions Green eoe oteAdS/SCFT the to devoted s tdtrsi h amplitudes the in terms ated ewe hs w sources two these between t eatnini oue on focused is attention he anpoete fthe of properties main ecito fthe of description a s tcr-nlsso the of re-analysis atic ROM2F/99/19 UNIVERSITA` DEGLI STUDI DI ROMA “TOR VERGATA”
FACOLTA` DI SCIENZE MATEMATICHE, FISICHE E NATURALI Dipartimento di Fisica
=4 supersymmetric Yang–Mills theory N and the AdS/SCFT correspondence
Tesi di dottorato di ricerca in Fisica presentata da Stefano Kovacs
Relatori Professor Giancarlo Rossi Professor Massimo Bianchi
Coordinatore del dottorato Professor Piergiorgio Picozza
Ciclo XI
Anno Accademico 1997-1998 Contents
Acknowledgments v
Note added v
1 Rigid supersymmetric theories in four dimensions 1 1.1 General supersymmetry algebra in four dimensions ...... 2 1.2 Representations of the supersymmetry algebra ...... 5 1.3 =1superspace ...... 10 N 1.4 =1supersymmetricmodels ...... 17 N 1.4.1 Supersymmetric chiral theories: Wess–Zumino model . .... 18 1.4.2 Supersymmetricgaugetheories ...... 19 1.5 Theories of extended supersymmetry ...... 22 1.5.1 Extended superspace: a brief survey ...... 23 1.5.2 > 1supersymmetricmodels...... 27 N 1.6 Quantization of supersymmetric theories ...... 29 1.6.1 Generalformalism ...... 30 1.6.2 Superfieldpropagators ...... 33 1.6.3 Feynmanrules...... 35 1.7 General properties of supersymmetric theories ...... 36
2 =4 supersymmetric Yang–Mills theory 40 N 2.1 Diverse formulations of =4 supersymmetric Yang–Mills theory . . . 41 N 2.2 Classical symmetries and conserved currents ...... 46 2.3 Finiteness of =4 supersymmetric Yang–Mills theory ...... 50 N 2.4 S-duality...... 52 2.4.1 Montonen–Olive and SL(2,Z)duality ...... 54 2.4.2 S-duality in =4 supersymmetric Yang–Mills theory . . . . . 57 N ii Table of contents iii
2.4.3 Semiclassical quantization ...... 60
3 Perturbative analysis of =4 supersymmetric Yang–Mills theory 66 N 3.1 Perturbation theory in components: problems with the Wess–Zumino gauge...... 68 3.1.1 One loop corrections to the propagator of the fermions belong- ingtothechiralmultiplet ...... 72 3.1.2 One loop corrections to the propagator of the scalars belonging tothechiralmultiplet ...... 74 3.2 Perturbation theory in =1superspace: propagators ...... 79 N 3.2.1 Propagatorofthechiralsuperfield...... 81 3.2.2 Propagatorofthevectorsuperfield ...... 86 3.2.3 Discussion...... 93 3.3 Three- and four-point functions of =1superfields ...... 96 N 3.3.1 Three-pointfunctions...... 96 3.3.2 Four-pointfunctions ...... 99
4 Nonperturbative effects in =4 super Yang–Mills theory 108 N 4.1 Instanton calculus in non Abelian gauge theories ...... 110 4.1.1 Yang–Millsinstantons ...... 111 4.1.2 Semiclassical quantization ...... 114 4.1.3 Bosoniczero-modes...... 116 4.2 Instanton calculus in the presence of fermionicfields ...... 118 4.2.1 Semiclassical approximation ...... 119 4.2.2 Fermioniczero-modes...... 121 4.3 Instanton calculus in supersymmetric gaugetheories...... 123 4.4 Instanton calculations in =4 supersymmetric Yang–Mills theory . . 125 N 4.5 A correlation function of four scalar supercurrents ...... 126 4.6 Eight- and sixteen-point correlation functions of current bilinears . . 136 4.6.1 The correlation function of sixteen fermionic currents ..... 136 4.6.2 The correlation function of eight gaugino bilinears ...... 137 4.7 Generalization to arbitrary N and to any K for large N ...... 138 iv Table of contents
5 AdS/SCFT correspondence 141 5.1 Type IIB superstring theory: a bird’s eye view ...... 142 5.2 D-branes...... 147 5.3 The AdS/SCFT correspondence conjecture ...... 151 5.4 Calculations of two- and three-point functions ...... 160 5.5 Four-pointfunctions ...... 170 5.6 D-instanton effects in d=10 type IIB superstring theory ...... 177 5.6.1 Testing the AdS/CFT correspondence: type IIB D-instantons vs. Yang–Millsinstantons ...... 181 5.7 The D-instanton solution in AdS S5 ...... 184 5 × Conclusions 189
A Notations and spinor algebra 193
B Electric-magnetic duality 198 B.1 Electric-magnetic duality and the Dirac monopole ...... 198 B.2 Non Abelian case: the ’t Hooft–Polyakov monopole ...... 200 B.3 Monopolesandfermions ...... 204
Bibliography 207 Acknowledgments
This dissertation is based on research done at the Department of Physics of the Uni- versit`adi Roma “Tor Vergata”, during the period November 1995 - October 1998, under the supervision of Prof. Massimo Bianchi and Prof. Giancarlo Rossi. I would like to thank them for introducing me to the arguments covered in this thesis, for all what they have taught me and for their willingness and constant encouragement. I also wish to thank Prof. Michael B. Green for the very enjoyable and fruitful col- laboration on the subjects reported in the last two chapters. I am grateful to Yassen Stanev for many useful and interesting discussions. Finally I thank the whole theo- retical physics group at the Department of Physics of the Universit`adi Roma “Tor Vergata” for the very stimulating environment I have found there.
Note added
This thesis was presented to the “Universit`adi Roma Tor Vergata” in candidacy for the degree of “Dottore di Ricerca in Fisica” and it was defended in April 1999. Hence more recent developments are not discussed here and moreover the references are not updated. However I cannot help mentioning at least the very detailed and comprehensive review by O. Arhony, S. Gubser, J. Maldacena, H. Ooguri and Y. Oz on “Large N Field Theories, String Theory and Gravity” (hep-th/9905111), that is not cited in the bibliography. Actually this paper would have made my writing of the last chapter much simpler if it came out earlier! There are some direct developments of this thesis that have been carried on during the last few months. In particular the analysis of logarithmic singularities in four- point Green functions and their relation to anomalous dimensions of un-protected operators in =4 SYM has been finalised in collaboration with Massimo Bianchi, N Giancarlo Rossi and Yassen Stanev. This issue is only briefly addressed in the con- cluding remarks. The interested reader is referred to the paper “On the logarithmic behaviour in =4 SYM theory”, hep-th/9906188. N
v Chapter 1
Rigid supersymmetric theories in four dimensions
Supersymmetry was introduced in [1] as a generalization of Poincar´einvariance and then implemented in the context of four dimensional quantum field theory in [2]. The idea of a symmetry transformation exchanging bosonic and fermionic degrees of freedom has proved extremely rich of consequences. Although up to now there exists no experimental evidence for supersymmetry, it plays a central rˆole in theoretical physics. At the present day the interest in the study of supersymmetric theories relies on one side on the rˆole it plays in the construction of a more fundamental model beyond the standard model of particle physics (SM) and on the other on the possibility of deriving results that can hopefully be extended to phenomenologically more relevant models. The standard model describes the strong and electro-weak interactions within the framework of a SU(3) SU(2) U(1) non-Abelian gauge theory; despite the full × × agreement with experimental data there are various motivations for believing that it cannot be a truly fundamental theory. A first problem arises in relation with the grand unification of particle interactions. The gauge couplings are believed to be unified at an energy scale of the order of 1016 GeV. The standard model contains many free parameters (masses and mixing angles) that must be “fine-tuned” in order to achieve the grand unification (hierarchy problem): the problem follows from the
1 2 Chapter 1. Rigid 4d supersymmetry
existence of ultra-violet quadratic divergences and is solved in supersymmetric models in which the divergences, as will be discussed later, are milder. Moreover the standard model does not include gravitational interactions: at present the only consistent description of gravity at the quantum level is provided by (super)string theories and supersymmetry appears to be a fundamental ingredient of such theories. In a more immediate perspective supersymmetric theories can be of help in un- derstanding aspects of phenomenologically interesting models. For example a de- scription of confinement in asymptotically free theories as dual superconductivity has been proposed and various realizations of this mechanism have been observed in supersymmetric models. More recently a correspondence between four dimensional quantum field theories and supergravity has been suggested; this particular subject appears very promising and will be extensively studied in chapter 5. This first chapter contains a review of rigid supersymmetry in four dimensions. The material presented is by no means original, the aim is to establish the notations and to recall ideas and results, mainly about the quantization of supersymmetric theories, that will be extensively used in the following. In the first part of the chapter the supersymmetry algebra and its realizations are described. Then the formalism of superspace is introduced in somewhat more detail both at the classical and at the quantum level. The last section reviews general properties of supersymmetric models. An introduction to supersymmetry is provided by the books by P. West [3] and J. Wess and J. Bagger [4] (whose notations will be followed in this thesis) and by the review papers [5, 6]; superspace techniques are treated in more detail in [7]; for phenomenological aspects of supersymmetry see [8]; an extensive introduction to superstring theory is given in the book by M.B. Green, J.H. Schwarz and E. Witten [9].
1.1 General supersymmetry algebra in four dimen- sions
Quantum field theory models of interest for the description of fundamental inter- actions combine the Poincar´einvariance with the invariance under a global internal symmetry group G. In [10] Coleman and Mandula proved, under general assumptions 1.1 Supersymmetry algebra 3
following from the axioms of quantum field theory, a “no-go” theorem that, states that the most general symmetry group for a theory with non trivial S-matrix is the direct product of the Poincar´egroup and an internal group G, which must be of the form of a semisimple Lie group times possible U(1) factors. Historically supersymmetry was discovered as an attempt to evade the constraints of the Coleman–Mandula no-go theorem .
The Poincar´egroup P=ISO(3,1) is the semidirect product of the group T4 of four dimensional translations and the Lorentz group SO(3,1); it is generated by Pµ, Mµν satisfying:
[Pµ,Pν]=0 [P , M ]=(η P η P ) (1.1) λ µν λµ ν − λν µ [M , M ]= (η M + η M η M η M ) . µν ρσ − µρ νσ νσ µρ − µσ νρ − νρ µσ
(See appendix A for the conventions and a collection of useful formulas). The generic internal symmetry group is a Lie group defined by the commutation relations:
a b ab c [T , T ]= if cT . (1.2)
Internal symmetries are supplemented by the discrete ones C, P and T. The supersymmetry algebra is a graded extension of (1.1): the theorem of Cole- man and Mandula is bypassed by allowing anticommuting as well as commuting generators. In [11] the most general algebra allowed by this weakening of the hy- pothesis of [10] was constructed: it contains, beyond (1.1) and (1.2), anticommuting i 1 1 generators Qα, Qαj˙ that are Weyl spinors transforming in the ( 2 , 0) and (0, 2 ) repre- sentations of the Lorentz group respectively, so that they do not commute with the generators Pµ and Mµν . Here α, α˙ =1, 2 are spinor indices and i, j ranging from 1 to 1 label the various of supersymmetries. N ≥ The most general symmetry algebra of a supersymmetric quantum field theory is therefore given by (1.1) and (1.2) supplemented by
j Qi , Q =2σµ P δij { α β} αα˙ µ i [Pµ, Qα] = [Pµ, Qαj˙ ]=0 i β i [Qα, Mµν]=(σµν )α Qβ 4 Chapter 1. Rigid 4d supersymmetry
˙ α˙ β [Qαj˙ , Mµν ]=(σµν ) β˙ Qj Qi , Qj = ε Zij { α β} αβ Q , Q ˙ = ε ˙ Z† (1.3) { αi˙ βj} α˙ β ij i a ai j [Qα, T ]= B jQα a ai [T , Qαi˙ ]= Bj† Qαi˙ ij [Z ,X] = [Zlk† ,X]=0 ,
aj aj where Bi† = B i. In (1.3) X denotes an arbitrary generator in the algebra so that Zij = Zji (central charges) generate the center of the algebra 1. Central charges − play a major rˆole in the quantization of -extended supersymmetric theories. They N generate an Abelian subalgebra so that
ij ij b Z = ab T ,
with Bi abjk = ab ijB k. aj − aj The supersymmetry algebra (1.3) considered as a graded Lie algebra possesses a group of automorphisms which is U( ) in the general case . The U( ) group of N N i automorphisms acts on the charges Qα and Qαj˙ as
i i k k Q U Q Q Q U † . α −→ k α αj˙ −→ αk˙ j In the particular case of =1 supersymmetry the U(1) group of automorphisms is N generated by R such that
[Q , R]= Q , [Q , R]= Q . α α α˙ − α˙
i 1 The supersymmetry charges Qα and Qαj˙ transform in spin 2 representations of the i Lorentz group; this implies that acting with Qα or Qαj˙ on a state of spin j produces a state of spin j 1 : supersymmetry generators exchange bosonic and fermionic ± 2 states. Reversing the line of reasoning and starting with a symmetry transforming the fields among themselves in such a way as to mix bosons and fermions, one concludes that the different physical dimension of bosonic and fermionic fields implies that the transformation must involve derivatives, i.e. space-time translations. This gives an 1In principle the antisymmetry of Zij is expected to imply that no central extension can exist in non extended ( =1) theories. It can be proved [12] that a central extension can actually be present in =1 supersymmetricN Yang–Mills theory. See [13] for a recent discussion in the context of the AdS/CFTN correspondence. 1.2 Representations of supersymmetry 5
intuitive explanation of how the requirement of a non trivial mixing of internal and space-time symmetries naturally leads to supersymmetry. An immediate consequence of the supersymmetry algebra is that in supersym- metric theories the energy is positive definite and vanishes only on supersymmetric (ground) states. The first equation in (1.3) implies
i i 0 Q , (Q )† + Q , (Q )† = 4 P =4 H , ≤ { 1 1i } { 2 2i } − N 0 N i X so that
ψ H ψ 0 h | | i ≥ for every state ψ and H Ψ = 0 if and only if Q ψ = 0, i.e. if ψ is supersymmetric. | i | i | i | i 1.2 Representations of the supersymmetry alge- bra
The irreducible representations of the general supersymmetry algebra can be con- structed starting from the (anti)commutation relations of the preceding section and the definition of the Casimir operators. For the Poincar´egroup the quadratic Casimir 2 µ 2 µ operators are P = PµP and W = WµW , where Wµ is the Pauli–Lubanski vector 1 W = ε P νM ρσ . µ 2 µνρσ 2 Here W is not a Casimir anymore, because Mµν does not commute with the su- i 2 persymmetry generators Qα and Qαj˙ , and is substituted by C defined (for the case =1) by N 2 µν C = Cµν C C = B P B P µν µ ν − ν µ 1 B = W Q σαα˙ Q . µ µ − 4 α˙ µ α The irreducible representations can be constructed by Wigner’s technique of induced representations. A general result following from the supersymmetry algebra is that every irre- ducible representation contains an equal number of bosonic and fermionic states. Defining the fermion number operator ( )Nf such that − ( )Nf B =+ B , ( )Nf F = F , − | i | i − | i −| i 6 Chapter 1. Rigid 4d supersymmetry
where B denotes a bosonic state and F a fermionic one, it follows that | i | i ( )Nf Q = Q ( )Nf , ( )Nf Q = Q ( )Nf . − α − α − − α˙ − α˙ − These relations imply
0=tr ( )Nf Q , Q =2σµ δijP tr( )Nf , − { α α˙ } αα˙ µ − so that in conclusion for non vanishing Pµ
tr( )Nf =0 − from which the statement follows.
Massive representations without central charges.
For particles of mass M in the rest frame P = ( M, 0, 0, 0) the supersymmetry µ − algebra reduces to
i i Q , Q ˙ =2Mδ δ { α βj} αα˙ j Q , Q = Q Q ˙ =0 , (1.4) { α β} { αi˙ βj} and creation and annihilation operators can be defined as follows 1 ai = Qi α √2M α 1 ai † = Q . (1.5) α √2M αi˙ i i aα and (aα)† satisfy the algebra of fermionic creation and annihilation operators. In fact by direct calculation one finds
ai , aj † = δβδi { α β } α j j j a i , a = ai † , a † =0 . (1.6) { α β} { α β } Given a Clifford vacuum Ω defined by | i ai Ω =0 i, α , α| i ∀ with P 2 Ω = M 2 Ω , the states are constructed by acting with the operators (ai )† | i − | i α on Ω : | i (n) α ...α 1 Ω 1 n = ain † . . . ai1 † Ω . (1.7) | i1...in i √n! αn α1 | i 1.2 Representations of supersymmetry 7
Antisymmetry under the exchange of pairs (αk, ik), (αl, il) implies n 2 . For each 2 ≤ N n the number of states (degeneracy) is , so that the total number of states nN (i.e. the dimension of the representation) is
2 N 2 2 d = N =2 N . n n=0 X 2 1 2 1 The representation contains 2 N− bosonic and 2 N− fermionic states. The highest spin state is obtained symmetrizing the maximum number of spinor indices, namely , leading to the value 1 for the highest spin in the multiplet. N 2 N
Massless representations without central charges.
In the “light-like” reference frame P =( E, 0, 0, E), with P 2 = 0, the supersymme- µ − try algebra reads
i 2E 0 i Q , Q ˙ = 2 δ { α βj} 0 0 j i j Q , Q = Q , Q ˙ =0 . { α β} { αi˙ βj} The natural definition of creation and annihilation operators is then 1 ai = Qi 2√E 1 i 1 i a† = (a )† = Q˙ . (1.8) i 2√E 1 They generate the Clifford algebra
i i a , a† = δ { j} j i j a , a = a†, a† =0 (1.9) { } { i j} i Notice that in this case only half of the generators Qα can be used to construct the i fermionic oscillators because Q2 and Q2˙j anticommute. Starting with a state of lowest helicity λ, Ω , playing the rˆole of a Clifford vacuum, defined by | λi ai Ω =0 | λi the states which constitute the multiplet are constructed as
(n) 1 n † † Ωλ+ ;i ,...,i = ain ...ai Ωλ . (1.10) | 2 1 n i √n! 1 | i 8 Chapter 1. Rigid 4d supersymmetry
(n) n n The state Ωλ+ ;i ,...,in has helicity λ + 2 and because of antisymmetry in the indices | 2 1 i
i1,... ,in is N times degenerate. The highest helicity in the multiplet is λ = n λ + N2 . The dimension of the representation given by the total number of states is
N d = =2N . Nn i=0 X CPT invariance implies in general a doubling of the number of states because it re- verses the helicity. So for example the =3 multiplet in four dimensions coincides N with the =4 multiplet if CPT invariance is required. Multiplets of particular rel- N evance such as =2 with λ = 1 , =4 with λ = 1 and =8 with λ = 2 are N − 2 N − N − automatically CPT invariant.
Supersymmetry representations in the presence of central charges.
For P 2 = M 2 in the rest frame P =( M, 0, 0, 0) the supersymmetry algebra with − µ − non vanishing central charges is given by
Qi , Q = 2Mδi δ { α αj˙ } j αα˙ Qi , Qj = ε Zij (1.11) { α β} αβ Q , Q ˙ = ε ˙ Z† , { αi˙ βj} − α˙ β ij with Zij = Zji, Z = Zij. Since Zij commute with all the generators, by invoking − ij − Schuur’s lemma there exists a basis in which they are diagonal with eigenvalues Zij. The antisymmetric matrix with elements Zij can be put in the following form N×N ij i kl T j Z = U kZ˜ (U )l ,
with U is a unitary matrix and
Z˜ = ε D (for even) ⊗ N ε D 0 Z˜ = ⊗ (for odd) , 0 0 N 0 1 where ε = and D is a diagonal N N matrix with eigenvalues Z , 1 0 2 × 2 R − such that Z∗ = Z , Z > 0. Considering the case with even the algebra can be R R R N rewritten in a more suitable form by decomposing the indices i and j as
i =(a, R) , j =(b, S) , 1.2 Representations of supersymmetry 9
with a, b =1, 2 and R,S =1,..., N2 . Now the algebra can be put in the form
aR bS ab RS Q , Q ˙ = 2Mδ δ δ { α β } αα˙ QaR, QbS = ε εabδRS Z (1.12) { α β } αβ S aR bS ab RS Q , Q ˙ = ε ˙ ε δ Z , { α˙ β } α˙ β S so that creation and annihilation operators can be defined in the following way:
R 1 1R 2R R R a = Q + εαβ(Q )† , (a )† = a† α √2 α β α α R 1 1R 2R R R b = Q εαβ(Q )† , (b )† = b† α √2 α − β α α These operators satisfy the algebra of fermionic creation and annihilation operators
aR, aS = bR, bS = aR, bS =0 { α β } { α β } { α β } R S RS a , (a )† = δ δ (2M + Z ) (1.13) { α β } αβ S R S RS b , (b )† = δ δ (2M Z ) . { α β } αβ − S
Since A, A† is a positive definite operator A, it immediately follows that Z 2M. { } ∀ R ≤ For Z < 2M R the structure of the multiplet is the same as in the case of no central R ∀ charges. For ZR =2M with R =1,..., R a subset of operators b cancels out and the algebra reduces to a 2( R) dimensional Clifford algebra giving rise to so called N − short multiplets. States belonging to short multiplets, also referred to as Bogomol’nyi Prasad Sommerfield (BPS) saturated states, play a crucial rˆole in the non perturbative dynamics of supersymmetric gauge theories as will be discussed later. As an example of the previous construction consider the simplest case, namely =1 massive multiplet with a spin zero Clifford vacuum Ω : the states are N | i Ω (spin 0) | i 1 (a )† Ω (spin ) α | i 2 1 1 γ (a )†(a )† Ω = ε (a )†(a )† Ω (spin 0) √2 α β | i − 2√2 αβ γ | i In the following the discussion will focus mainly on the =4 massless case. For N a rigid supersymmetric theory with fields of spin not larger than one the multiplet contains the following states with relative degeneracies 10 Chapter 1. Rigid 4d supersymmetry
helicity 1 1 0 1 1 − − 2 2 degeneracy 1 4 6 4 1
Given the irreducible representations of the supersymmetry algebra the corresponding realizations in terms of fields can be immediately deduced. Although this was the historical pattern in the construction of supersymmetric quantum field theory models, it appears more natural to derive the structure of supermultiplets using the superfield formalism, so this will be the approach followed here. Notice that Wigner’s method here employed gives representations of “on-shell” states. It is always possible to construct sets of fields realizing such irreducible rep- resentations, but on the contrary it is usually not possible, for > 1, to achieve the N closure of the supersymmetry algebra on a set of “off-shell” fields, i.e. without use of the equations of motion. The irreducible representations of supersymmetry have automatically a well de- fined transformation under the group of automorphisms of the algebra (1.3). For massless representations the largest subgroup of the automorphisms group that re- spects helicity is the whole U( ), whereas for massive states it depends on the N presence and number of central charges. In the absence of central charges, or if none of them saturates the BPS bound, the spin preserving automorphisms group is USp(2 ). It reduces to USp( ) for even and to USp( + 1) for odd if one N N N N N central charge satisfies the condition Z =2M.
1.3 =1 superspace N Ordinary quantum field theories are constructed in terms of field operators that are functions (actually distributions) defined on Minkowskian space-time, parametrized
by the coordinates xµ. The action is required to be invariant under Poincar´espace- time symmetry as well as under transformations of an internal group G. The trans- formation of a generic field f(x) is of the form
R(g)f(x)= f(τ(g)x) , (1.14)
with R(g) and τ(g) suitable representations of the relevant symmetry group. In supersymmetric models the space-time symmetry is extended by the intro- i duction of the transformations generated by anticommuting charges Qα, Qαj˙ . The 1.3 Superspace 11
natural way to implement the extended symmetry algebra is to construct the theory in terms of generalized fields, called superfields [14, 15], depending on the coordinates xµ as well as on “fermionic” coordinates θα and θα˙
F = F (xµ, θα, θα˙ ) . (1.15)
Starting from the supersymmetry algebra a generic element of the corresponding group of transformations can be written in the form
µ α α˙ g(x, θ, θ)= ei(x Pµ+θ Qα+θα˙ Q ) .
The action on superfields is then defined through a differential representation of the α˙ operators Pµ, Qα and Q . This construction can be carried out in a more general fashion, suitable for generalization to the case of -extended supersymmetry, defining the superspace N parametrized by
zm =(xµ, θα, θα˙ ) as a coset space. =1 superspace under consideration can be obtained as the coset N SP/L, where SP is the super Poincar´egroup generated by (1.3) and L is the Lorentz group SO(3,1). A coset group K=G/H can be parametrized by coordinates ξ , m =1,..., (dim G m − dim H). By writing an element of G through the exponential map
m k g = eiξ Lm eiζ Hk , where the generators of G split into the generators of H (Hk) and the remaining Ln, elements of the coset are obtained by taking ζk = 0. For the present case K=SP/L this implies that one should write
α˙ i(aµP +ηαQ +η Q ) 1 wµν M g = e µ α α˙ e 2 µν (1.16) so that wµν = 0 gives the elements of K
µ α α˙ m k = ei(x Pµ+θ Qα+θα˙ Q ) = ez Km which are parametrized by z =(x , θ , θ ); as a result =1 superspace is an eight m µ α α˙ N dimensional manifold. Under the action of a group element g0 =(aµ, ηα, ηα˙ )
m ′m 1 ′µν z Km z Km w Mµν g e = e e 2 g′ . 0 ◦ ≡ 12 Chapter 1. Rigid 4d supersymmetry
An explicit calculation using Becker–Hausdorff’s formula and anticommutation of Q
and P gives the following transformations of the coordinates zm
µ µ µ α µ α˙ α µ α˙ µν x′ = x + a + iη σ θ iθ σ η + w x αα˙ − αα˙ ν α α α 1 µν α θ′ = θ + η + w σ θ (1.17) 4 µν β β α˙ α˙ 1 β˙ θ′ = θ + ηα˙ θ σµν α˙ w . − 4 β˙ µν
m m Now comparison of (g g ) ez Km with g (g ez Km ) allows one to prove that 1 ◦ 2 ◦ 1 ◦ 2 ◦ the generators Q and Q satisfy the supersymmetry algebra (1.3) for =1. α α˙ N As already mentioned a superfield is a function of zm
F (z)= F (x, θ, θ) .
The transformation under the group action is the generalization of (1.14) that imple-
ments z′ = τ(g)z with z′ given by (1.17). For infinitesimal transformations
δ F = F (z + δz) F (z)= δg XmF (z) , g − m which is achieved realizing the super Poincar´egenerators through the differential
operators Xm =(ℓm,ℓα, ℓα˙ ,ℓµν )
ℓµ = ∂µ ∂ α˙ ℓ = + iσµ θ ∂ α ∂θα αα˙ µ ∂ α µ ℓα˙ = α˙ + iθ σαα˙ ∂µ (1.18) ∂θ ˙ 1 β µν α ∂ 1 β µνα˙ ∂ ℓµν = (xµ∂ν xν ∂µ) θ σβ + θ σ ˙ α˙ . − − − 2 ∂θα 2 β ∂θ A calculation shows that they satisfy the super Poincar´ealgebra
ℓ ,ℓ =0 , ℓ , ℓ =2iσµ ℓ , { α β} { α α˙ } αα˙ µ (1.19) 1 [ℓ ,ℓ ]=0 , [ℓ ,ℓ ]= σ βℓ . µ α α µν 2 µνα β Superfields defined in this way form linear representations of supersymmetry. Since supersymmetry acts on superfields through linear differential operators it follows that the product of superfields is a new superfield. 1.3 Superspace 13
The objects considered up to now are scalar superfields 2; more general superfields transforming in a non trivial way under Lorentz transformations can be considered as well. In general a superfield carrying a set of Lorentz indices encoded in a single label p transforms as
1 µν q w Mµν R(g)fp(z)= Dp e− 2 Fq(τ(g)z) , q where Dp is a suitable representation of the Lorentz group. Physical fields directly related to the states considered in section 1.2 can be ob- tained from the superfields by expanding in the θ and θ variables. Since θα and θα˙ are Weyl spinors the expansion reduces to a polynomial. For a generic superfield
α α˙ α α˙ F (x, θ, θ) = f(x)+ θ χα(x)+ θα˙ ψ (x)+ θ θαg(x)+ θα˙ θ h(x)+ α µ α˙ α α˙ α˙ α α α˙ +θ σαα˙ θ rµ(x)+ θ θαθα˙ λ (x)+ θα˙ θ θ ξα(x)+ θ θαθα˙ θ s(x) . (1.20)
The fields f(x), χ(x), ψ(x), g(x), h(x), rµ(x), λ(x), ξ(x) and s(x) constitute a supermultiplet and their transformation law is derived from the transformation of the superfield by defining
α α˙ α α˙ δηF (x, θ, θ)= δηf(x)+ θ δηχα(x)+ θα˙ δηψ (x)+ θ θαδηg(x)+ θα˙ θ δηh(x)+ α µ α˙ µ α α˙ α˙ α α α˙ +θ σαα˙ θ δηr (x)+ θ θαθα˙ δηλ (x)+ θα˙ θ θ δηξα(x)+ θ θαθα˙ θ δηs(x) (1.21) and matching terms with the same powers of θ and θ. Linear representations of supersymmetry constructed in this way are in general reducible; the irreducible representations studied in section 1.2 are in correspondence with constrained superfields. All the representations of rigid =1 supersymmetry in N four dimensions are realized in terms of two basic superfields, namely chiral superfields and vector superfields.
From now on the differential operators ℓα and ℓα˙ will be denoted as Qα and Qα˙ like the abstract generators since they satisfy the same algebra. The operators Qα and Qα˙ realize the left regular representation of supersymmetry on superfields, it is useful to analogously define the right regular representation realized by the differential operators
∂ µ α˙ ∂ α µ Dα = + iσαα˙ θ ∂µ , Dα˙ = α˙ iθ σαα˙ ∂µ (1.22) ∂θα −∂θ − 2From now on scalar superfields will be referred to simply as superfields. 14 Chapter 1. Rigid 4d supersymmetry
satisfying
D , D = 2iσµ ∂ { α α˙ } − αα˙ µ D ,D = D , D ˙ =0 . { α β} { α˙ β}
Dα and Dα˙ are usually referred to as superspace covariant derivatives, actually they can be obtained from the general expression for the covariant derivative in the coset
space SP/L: the fact that Dα and Dα˙ do not reduce to ∂α and ∂α˙ despite the flatness of the =1 superspace is a consequence of the presence of a non vanishing torsion. N
Chiral superfields.
Chiral superfields are characterized by the constraint
Dα˙ Φ(x, θ, θ)=0 ,
which takes a particularly simple form expressing Φ(x, θ, θ), D and D in terms of the variable yµ = xµ + iθσµθ. As functions of y, θ and θ one gets
∂ α˙ ∂ D = +2iσµ θ α ∂θα αα˙ ∂yµ ∂ Dα˙ = α˙ , −∂θ so that the general chiral superfield assumes the form
Φ(y, θ)= φ(y)+ √2θψ(y)+ θθF (y) .
Substituting yµ = xµ + iθσµθ and expanding in θ and θ yields
µ Φ(x, θ, θ) = φ(x)+ √2θψ(x)+ iθσ θ∂µφ(x)+
i µ 1 +θθF (x) θθ∂µψ(x)σ θ + θθθθ2φ(x) . (1.23) − √2 4 Analogously antichiral superfields are defined by
DαΦ†(x, θ, θ)=0 .
µ µ µ µ The natural variable is y† = x iθσ θ; in terms of y† −
Φ† = Φ†(y†, θ)= φ∗(y†)+ √2θψ(y†)+ θθF ∗(y†) . 1.3 Superspace 15
The field components associated to chiral and antichiral multiplets are of two complex scalars (φ, F ) and one Weyl spinor (ψ) corresponding to twice the number of states in the massive on-shell =1 multiplet of section 1.2. N Products of (anti) chiral superfields are still (anti) chiral superfields. For example
ΦiΦj = φi(y)φj(y)+ √2θ[ψi(y)φj(y)+ φi(y)ψj(y)]+ +θθ[φ (y)F (y)+ φ (y)F (y) ψ (y)ψ (y)] i j j i − i j 3 3 The product Φ†Φ is neither chiral nor antichiral. Moreover since D = D = 0 it immediately follows that for a generic superfield F
2 D F = Φ is chiral 2 D F = Φ† is antichiral .
Notice that the double constraint DαΦ= Dα˙ Φ = 0 implies that Φ is constant.
Vector superfields.
Vector superfields obey the constraint
V †(x, θ, θ)= V (x, θ, θ) , which gives rise to the component expansion i V (x, θ, θ)= C(x)+ iθχ(x) iθχ(x)+ θθS(x)+ − √2
i µ i µ θθS†(x) θσ θA (x)+ iθθθ λ(x)+ σ ∂ χ(x) + −√ − µ 2 µ 2 i 1 1 iθθθ λ(x)+ σµ∂ χ(x) + θθθθ D(x)+ 2C(x) . (1.24) − 2 µ 2 2 The supermultiplet contains two real scalars (C, D), one complex scalar (S), four
Weyl spinors (χ, χ, λ, λ) and one real vector (Aµ). This is twice the content of the massive =1 multiplet obtained starting with a Clifford vacuum of spin 1 and also N 2 of the massless CPT invariant multiplet with minimum helicity λ = 1. − Since the vector multiplet contains a vector field Aµ(x) it is the basic ingredient for the construction of supersymmetric gauge theories. The supersymmetric general- ization of an Abelian gauge transformation [16, 17] is
V V +Φ+Φ† , −→ 16 Chapter 1. Rigid 4d supersymmetry
with
µ Φ + Φ† = φ + φ∗ + √2(θψ + θψ)+ θθF + θθF ∗ + iθσ θ∂ (φ φ∗)+ µ − 1 µ 1 µ 1 + θθθσ ∂µψ + θθσ θ∂µψ + θθθθ2(φ + φ∗) . √2 √2 4 For the components one gets
C C + φ + φ∗ −→ χ χ i√2ψ −→ − S S i√2F −→ − A A i∂ (φ φ∗) µ −→ µ − µ − λ λ −→ D D . −→ Notice that the transformations for the lowest components C, χ and S are purely algebraic (no derivatives involved), so they can always be used to put these fields to zero. This particular choice of gauge is known as Wess–Zumino (WZ) gauge. Subtleties related to the choice of the Wess–Zumino gauge will be discussed in chapter 3. It is important to note that in this gauge one has 1 V (x, θ, θ) = θσµθA (x)+ iθθθλ(x) iθθθλ(x)+ θθθθD(x) − µ − 2 1 V 2(x, θ, θ) = θθθθA (x)Aµ(x) −2 µ V n(x, θ, θ) = 0 , n 3 . ∀ ≥ Fixing the Wess–Zumino gauge still leaves an arbitrary Abelian gauge freedom on
Aµ. From the vector superfield it is straightforward to construct a superfield containing
the field strength for the field Aµ: it is defined as 1 W (x, θ, θ)= DDD V (x, θ, θ) (1.25) α −4 α
and is a spinor superfield. Besides Wα it is useful to introduce 1 W (x, θ, θ)= DDD V (x, θ, θ) . (1.26) α −4 α˙ They are respectively chiral and antichiral:
Dα˙ Wα =0 ,DαW α˙ =0 1.4 Supersymmetric models 17
and gauge invariant
W [V +Φ+Φ†]= W [V ] .
Gauge invariance of Wα allows to compute its component expansion in the Wess– Zumino gauge; in terms of yµ one gets
i W (y, θ) = iλ(y)+ δ βD (σµσν ) β(∂ A (y) ∂ A (y)) θ + α − α − 2 α µ ν − ν µ β µ α˙ +θθσαα˙ ∂µλ (y)= (1.27) i α˙ = iλ(y)+ δ βD (σµσν) βF (y) θ + θθσµ ∂ λ (y) − α − 2 α µν β αα˙ µ and analogously for W α.
1.4 =1 supersymmetric models N The superfield formalism allows a straightforward construction of supersymmetric ac- tions. The supersymmetry transformation of the (θθ) θθ component (F -component) of a (anti) chiral superfield is a total derivative, as can be deduced by specializing (1.21) to the case of a chiral superfield, so that an action of the form 3
4 2 2 S = d x d θ Φ+ d θ Φ† Z Z Z is automatically supersymmetric for an arbitrary chiral Φ. The same situation holds for the θθθθ component (D-component) of a vector superfield yielding the supersymmetric expression
S = d4x d4θ V , Z Z for all V ’s such that V † = V . Requirement of invariance under the group of automorphisms allows to further restrict the possible couplings in the action for supersymmetric models; in this context the symmetry under transformations of the group of automorphisms is referred to as R-symmetry.
3The properties of Grassmannian integration and the conventions here employed are summarized in appendix A. 18 Chapter 1. Rigid 4d supersymmetry
1.4.1 Supersymmetric chiral theories: Wess–Zumino model
The action for the Wess–Zumino model [18] is constructed in terms of only chiral superfields and reads
4 4 2 1 S = d x d θ Φ†Φ + d θ m Φ Φ + i i 2 ij i j Z Z 1 + g Φ Φ Φ + λ Φ + h.c. . (1.28) 3 ijk i j k i i In terms of component fields it becomes
4 µ µ 1 S = d x ∂ φ∗∂ φ iψ σ ∂ ψ + F ∗F + m (φ F ψ ψ )+ − µ i i − i µ i i i ij i j − 2 i j Z +g (φ φ F ψ ψ φ )+ λ F + h.c. . (1.29) ijk i j k − i j k i i The action of the Wess–Zumino model is invariant under the supersymmetry trans- formations
δηφi = √2ηψi µ δηψi = i√2σ η∂µφi + √2ηFi µ δηFi = i√2ησ ∂µψi
The action (1.29) does not contain derivatives of the fields Fi. They are auxiliary fields: their equations of motion are algebraic and can be used to eliminate them
from the action. Solving the equations for Fi ∂ L = Fk + λk∗ + mik∗ φi∗ + gijk∗ φi∗φj∗ =0 ∂Fk∗ ∂ L = Fk∗ + λk + mikφi + gijkφiφj =0 , ∂Fk yields the Lagrangian
µ µ 1 1 = iψ σ ∂ ψ ∂ φ∗∂ φ m ψ ψ m∗ ψ ψ + L − i µ i − µ − 2 ij i j − 2 ij i j g ψ ψ φ g∗ ψ ψ φ∗ (φ ,φ∗) , − ijk i j k − ijk i j k − V i i where the potential is given (for λ = 0) by the expression V i
=(F ∗F )=(m φ + g φ φ )(m∗ φ∗ + g∗ φ∗ φ∗ ) . (1.30) V k k ik i ijk i j mk m mnk m n 1.4 Supersymmetric models 19
This scalar potential in particular contains a φ4 interaction and the mass terms for the scalars φk; actually the Wess–Zumino model is the most general renormalizable supersymmetric theory involving only chiral superfields. The action (1.29) is the most general renormalizable supersymmetric action that can be constructed in terms of only chiral superfields. More generally one can consider an invariant action of the form
4 4 2 S = d x d θ Φi†Φi + d θ W(Φ) + h.c. , Z Z Z where W(Φ) is a holomorphic function of Φ called superpotential. In this case the scalar potential becomes
∂W(Φ) 2 (Φ) = . V ∂Φ
The Wess–Zumino model can be constructed in terms of unconstrained superfields 3 using the property D3 = D = 0 and defining
2 2 Φ= D Ψ Φ† = D Ψ† ,
The unconstrained action that reduces to (1.28), for the case of one single field Φ, is
4 4 2 2 2 2 S = d xd θ (D Ψ†)(D Ψ) 2m(ΨD Ψ+Ψ†D Ψ+ − Z n 4g 2 2 2 2 Ψ(D Ψ) +Ψ†(D Ψ†) . (1.31) − 3! h i The equivalence follows using the identity
1 2 d4x d2θ Ψ= d4x d4θ D Ψ . −4 Z Z Z Z 1.4.2 Supersymmetric gauge theories
Gauge invariant interactions can be constructed in terms of vector superfields and more precisely in terms of the superfield Wα defined in section 1.3, which contains the gauge field strength among its components. Given a chiral superfield Φ and a vector one V an Abelian gauge theory is obtained as follows. The gauge transformation for the chiral superfield is
iqkΛ Φk′ = e− Φk , with Dα˙ Λ=0 † iqkΛ Φk† ′ = e Φk† , with DαΛ† =0 , 20 Chapter 1. Rigid 4d supersymmetry
where qk denotes the charge of the superfield Φk. The gauge parameter must be a chiral superfield Λ. Supplementing these transformations with the following one for the vector superfield V
V ′ = V + i(Λ Λ†) , − an invariant action can be written in the form
4 2 1 α 2 1 α˙ 4 q V S = d x d θ W W + d θ W W + d θ Φ†e i Φ + 4 α 4 α˙ i i Z Z Z Z 1 1 + d2θ m Φ Φ + g Φ Φ Φ + h.c. (1.32) 2 ij i j 3 ijk i j k Z Invariance under the above defined transformations is immediately verified. Notice that the action (1.32) is non-polynomial: this a general feature of supersymmetric gauge theories and is a consequence of the adimensionality of the vector superfield V . Choosing the Wess–Zumino gauge reduces (1.32) to a polynomial form since it implies V 3 = 0. This choice allows one to prove the renormalizability of the model. The vector superfield V contains the gauge dependent components C, χ and S that can be put to zero by choosing the Wess–Zumino gauge as well as the real scalar D. The latter is an auxiliary field since the corresponding equations of motion obtained from (1.32) are purely algebraic. This construction can be generalized to the non Abelian case [19, 20]. For a
generic non Abelian compact gauge group G the fields Φ, Φ† and V are matrices.
a a a † † (Φ)ij = TijΦa , Φ ij = TijΦa , Vij = TijVa , with generators T a of G satisfying
a b ab a b ab c tr(T T )= C(r)δ [T , T ]= if cT ,
C(r)= dr being the Dynkin index of the representation r. The gauge transformation takes the form
iΛ iΛ† Φ′ = e− Φ , Φ† ′ = Φ†e
and for the vector superfield
V ′ iΛ† V iΛ V V ′ : e = e− e e . (1.33) −→ 1.4 Supersymmetric models 21
It can be proved that in this case as well it is possible to choose the Wess–Zumino gauge in which V 3 = 0. For infinitesimal gauge transformations use of the Becker–Hausdorff’s formula allows to write
δV = V ′ V = i (Λ+Λ† + coth( )(Λ Λ†) , (1.34) − LV/2 LV/2 − where (B) = [A, B] is the Lie derivative and (1.34) is a compact form to be LA understood in terms of the power expansion of coth( ). L The non Abelian generalization of the field strength superfield Wα is
1 V V W = DDe− D e . α −4 α
In the Wess–Zumino gauge the non Abelian Wα has the same form as in (1.27) with F = ∂ A ∂ A + ig[A , A ] . µν µ ν − ν µ µ ν
The transformation law of Wα induced by (1.33) is easily obtained and reads
iΛ iΛ W W ′ = e− W e , α −→ α α so that while in the Abelian case Wα was gauge invariant for a non Abelian gauge group G it turns out to be covariant. The most general renormalizable action for interacting chiral and vector super- fields is therefore
4 1 1 2 α 2 α˙ 4 V V S = d x tr d θ W Wα + d θ W α˙ W + d θ e− Φ†e Φ+ dr 16g2 Z Z Z Z 1 1 + d2θ m Φ Φ + g Φ Φ Φ + h.c. . (1.35) 2 ij i j 3 ijk i j k Z The corresponding component field expansion in the Wess–Zumino gauge for the case mij =0, gijk = 0 is
4 1 a aµν a µ a 1 a a µ S = d x F F iλ σ λ + D D φ† φ+ −4 µν − Dµ 2 −Dµ D Z µ a a a a a a iψσ ψ + F †F + i√2g φ†T ψλ λ T φψ + gD φ†T φ , (1.36) − Dµ − where the correct factors of the coupling constant g have been recovered substituting in (1.35) V 2gV . In (1.36) denotes the ordinary covariant derivative −→ Dµ φ = ∂ φ + igAa T aφ Dµ µ µ ψ = ∂ ψ + igAa T aψ Dµ µ µ λa = ∂ λa + igf a Ab λc . Dµ µ bc µ 22 Chapter 1. Rigid 4d supersymmetry
Eliminating the auxiliary fields F and D through the equations of motion produces the scalar potential given by (1.30) plus the so called “D-term”
1 2 2 = g [φ†,φ] . VD 8 The action (1.36) can be proved to be invariant under the transformations
δξφ = √2ξψ δ ψ = i√2σµξ φ + √2ξF ξ Dµ µ a a δξF = i√2σ ξ µψ +2igT φξλ a D δ Aa = iλ σ ξ + iξσ λa (1.37) ξ µ − µ µ a µν a a δξλ = σ ξFµν + iξD a δ Da = ξσµ λ λaσµξ . ξ − Dµ −Dµ The Wess–Zumino condition is not supersymmetric so that it not only (partially) breaks gauge invariance, but it also breaks supersymmetry, that is recovered through a non linear realization given by equations (1.37). These transformations are more precisely a mixture of supersymmetry and the residual non Abelian gauge symmetry. The choice of the Wess–Zumino gauge leads to subtleties at the quantum level, as will be discussed in chapter 3. The super-gauge invariant action (1.35) can be generalized by introducing a com- plexified coupling constant θ 4πi τ = + 2π g2 obtaining, in the pure super Yang–Mills case, the analytic expression 1 S = Im τ d4xd2θ trW αW = 8π α Z 1 1 1 = d4x tr F F µν iλσµ λ + D2 + g2 −4 µν − Dµ 2 Z θ d4x trF F˜µν , −32π2 µν Z ˜µν 1 µνρσ where F = 2 ε Fρσ.
1.5 Theories of extended supersymmetry
The most straightforward way to construct multiplets for theories with extended ( > 1) supersymmetry is to assemble together =1 multiplets in such a way as to N N 1.5 Extended supersymmetry 23
obtain a field content corresponding to the representations of section 1.2. It turns out that in general this program can be achieved only on-shell; in other words it is not possible to realize -extended supersymmetry in a closed form through an off-shell N multiplet of fields. With the exception of =2 supersymmetric Yang–Mills theories N closure of the algebra without imposing the equations of motion would require an infinite set of auxiliary fields. This problem is related to the presence of central charges in the supersymmetry algebra for > 1. N As in the =1 case the notion of superspace will be introduced to start with and N then explicit realizations in specific models will be considered.
1.5.1 Extended superspace: a brief survey
The aim of this section is to give a brief account of ideas beyond the superspace formulation for theories with extended supersymmetry, no claim of completeness is made. In the following extended superspace will not be used in explicit calculations, but it will rather be invoked to justify some statements. A detailed and self-contained review of extended superspace is given in [21]. The general construction of superspace as a coset space introduced in section 1.3 can be generalized to define extended superspace as the coset K=SP∗/L, where SP∗ is the extended super Poincar´egroup generated by the general algebra (1.1), (1.3) and L is the Lorentz group SO(3,1). The generic element of SP∗ can be written
α˙ i(aµP +ηαQi +ηi Q +brZ + 1 w M µν ) g = e µ i α α˙ i r 2 µν ,
r where Zij = ΩijZr are the central charges of the supersymmetry algebra and in this case there are anticommuting generators Qi . Elements of the coset are obtained N α setting wµν = 0 and can be expressed in the form
µ α i i α˙ r k = ei(x Pµ+θi Qα+θα˙ Qi +ζ Zr) . (1.38)
Consequently they are parametrized by the enlarged set of coordinates
i j zm =(xµ, θα, θα˙ ,ζr) . (1.39)
Repeating the steps of section 1.3 allows to obtain the following supersymmetric 24 Chapter 1. Rigid 4d supersymmetry
coordinate transformations
µ µ α µ αi˙ α µ αi˙ x′ = x + iη σ θ iθ σ η i αα˙ − i αα˙ α α α θi′ = θi + ηi αj˙ αj˙ θ′ = θ + ηαj˙ (1.40) r r α r ij i αj˙ r ζ′ = ζ + iηi θαj(Ω ) + iηα˙ θ (Ω )ij .
µ i Under the action of the central charges x , θα and θαj˙ do not transform while
r r r ζ′ = ζ + b . (1.41)
Superfields are functions defined on the manifold parametrized by zm
i j F = F (xµ, θα, θα˙ ,ζr) . (1.42)
The action of the super Poincar´egroup on superfields is defined through
R(g) F (z)= F (z′)
where z′ is given in (1.40) and R(g) acts as the left regular representation. Differential operators generating the infinitesimal transformations now read
ℓµ = ∂µ
i ∂ µ αi˙ r ij ∂ ℓα = α + iσαα˙ θ ∂µ + i(Ω ) θαj r ∂θi ∂ζ ∂ α µ j r ∂ ℓαi˙ = αi˙ + iθi σαα˙ ∂µ + iθα(Ω )ij ∂θ ∂ζr ∂ ℓr = ∂ζr ˙ 1 β µν α ∂ 1 β µν α˙ ∂ ℓµν = (xµ∂ν xν ∂µ) θ σβ + θ σ ˙ α˙ − − − 2 ∂θα 2 β ∂θ ∂ αi˙ ∂ ℓ = θαBi θ Bi ∗ . a − i aj ∂θα − aj α˙ j ∂θj Relation to component field multiplets is established by Taylor expanding in the θ, θ as well as in the ζ coordinates. Notice that the ζ’s are “bosonic” coordinates, so that the expansion is a truly infinite series
(0) (1) (r) F (x, θ, θ,ζ)= F (x, θ, θ)+ F(r) (x, θ, θ) ζ + ... , (1.43) 1.5 Extended supersymmetry 25
where
F (0)(x, θ, θ) = F (x, θ, θ, 0)
(1) ∂ F(r) (x, θ, θ) = F (x, θ, θ,ζ) . ∂ζr ζ=0
“Super-covariant” derivatives generalizing those defined in section 1.3 are given by
Dµ = ∂µ
i ∂ µ αi˙ r ij ∂ Dα = α + iσαα˙ θ ∂µ i(Ω ) θαj r ∂θi − ∂ζ ∂ α µ r j ∂ Dαi˙ = αi˙ iθi σαα˙ ∂µ i(Ω )ijθα˙ (1.44) −∂θ − − ∂ζr ∂ D = . r ∂ζr
Just as in the case of =1 superspace constraints defining superfields correspond- N ing to irreducible representations of supersymmetry are implemented through the covariant derivatives (1.44). In general constrained superfields still depend on the coordinates ζ, so that the component supermultiplets contain an infinite number of fields; the only exception is the =2 vector superfield defined by the condition N ∂ D V (x, θ, θ,ζ)= V (x, θ, θ,ζ)=0 . (1.45) r ∂ζr
As will be discussed in the next subsection the component fields associated to this superfield are those defining the =2 supersymmetric Yang–Mills model. N > 1 irreducible on-shell representations constructed in section 1.2 contain a N 2 finite number of states (2 N in the massive case and 2N in the massless case). The reason for the appearance here of an infinite number of component fields is that closure of the supersymmetry algebra off-shell in the presence of central charges requires infinitely many auxiliary fields, the only exception being the =2 super N Yang–Mills multiplet related to the superfield in equation (1.45). The =2 supersymmetry algebra contains in principle two central charges, but N one of them can always be eliminated through a chiral rotation, so that the only relevant anticommuting relation is
Di ,Dj =2ε εijD . { α β} αβ ζ 26 Chapter 1. Rigid 4d supersymmetry
One can consider a superfield W satisfying
i Dα˙ W =0 ,DζW =0 . (1.46)
A multiplet of fields can be obtained taking the θ = 0 component of the fields
i αi j i W,DαW,D DαW,DαDβiW , i j k i j k l DαDβDγ W,DαDβDγ DδW . (1.47)
Exploiting the conditions (1.46) in (1.47) does not produce an irreducible multiplet. A further reduction is necessary which is achieved by imposing the reality condition
αi j i αj˙ D DαW = Dα˙ D W .
This gives rise to a multiplet containing the fields
i ij D , χα , C , Fµν , (1.48)
where D and Cij = Cji are real scalars, χi Weyl spinors and F = ∂ A ∂ A is α µν µ ν − ν µ the field strength for a real vector field A . These are the fields of the =2 super µ N Yang–Mills model: they sum up a chiral and a vector =1 multiplet. N An example of a =2 multiplet possessing central charge arises if one considers N a superfield satisfying 1 Di Φ = δi Dk Φ α j 2 j α k Dαi˙ Φj + Dαj˙ Φi =0 . (1.49)
Using these constraints a multiplet of fields can be obtained taking the θ = 0 com- ponent of the fields
k k Φ ,DαΦk , Dα˙ Φk ,DζΦ .
Some D-algebra manipulations allow to construct a supermultiplet containing two Weyl spinors and four complex scalars
ψα , χα , φi , Fi . (1.50)
This multiplet is known as hypermultiplet and can be viewed as the combination of two =1 chiral multiplets. It will be shown to describe =2 matter in supersymmetric N N gauge theories. 1.5 Extended supersymmetry 27
In general the constraints that need to be imposed on > 1 superfields cannot N be explicitly solved for the whole superfield; in other words considering all the θ- components yields a set of non independent fields. The lack of an explicit solution of the constraints does not allow to exploit the potentiality of the superspace formalism for extended supersymmetry, so that it does not prove so powerful a tool as in the =1 case. N These problems were partially solved in [22] by the introduction of harmonic su- perspace. It is based on a parametrization of =2 superspace as a homogeneous space N that leads to different bosonic coordinates associated to the central charges. Use of such coordinates allows to construct =2 hypermultiplets in terms of unconstrained N superfields. A generalization of this approach leading to the so called analytic super- space has been proposed in order to give a manifestly supersymmetric description of =4 theories (see the review [21] for the details), however it only works on-shell. N In the following all the explicit calculations using superfield formalism will be carried out within the framework of =1 superspace. N 1.5.2 > 1 supersymmetric models N In this section only examples of =2 theories will be discussed, the general properties N of =4 models, that are the main subject of this work, will be studied in detail in N next chapter.
=2 supersymmetric Yang–Mills model. N The superfield formulation of =2 pure supersymmetric Yang–Mills models is a N straightforward generalization of the = 1 case because it is constructed in terms N of a =2 superfield satisfying the constraint (1.45), which implies no dependence on N the central charges. Since no central charges are present a full off-shell formulation is possible. Superspace in this case is parametrized by
i ˜ ˜ zm =(xµ, θα, θαj˙ )=(xµ, θα, θα˙ , θα, θα˙ ) The starting point is a chiral superfield Ψ defined by
Dα˙ Ψ=0 , D˜ α˙ Ψ=0 (1.51)
There is a natural choice of variables to describe Ψ, namely ˜ ˜ y˜µ = xµ + iθσµθ + iθσµθ 28 Chapter 1. Rigid 4d supersymmetry
Ψ can be expressed in terms ofy ˜ as
α Ψ(˜y, θ) =Φ(˜y, θ)+ i√2θ˜ Wα(˜y, θ)+ θ˜θG˜ (˜y, θ) ,
where Φ and G are =1 chiral superfields and W is a chiral spinor superfield. N α To construct a non Abelian gauge theory Ψ is taken to be a matrix
a Ψij = TijΨa .
A =2 supersymmetric action can then be written as N 1 1 S = Im τ d4x d2θd2θ˜tr Ψ2 , (1.52) 4π 2 Z Z where
2 α Ψ = W W 2 + 2GΦ 2 . θ2θ˜2 α|θ |θ =2 super Yang–Mills is obtained by further constraining the chiral superfield G. N This is achieved by requiring Ψ to satisfy
αi j i αj˙ D DαΨ= Dα˙ D Ψ† .
Solving the constraint provides for G the following expression
2 V (˜y iθσθ,θ) G = d θ Φ†(˜y iθσθ, θ)e − . − Z Substituting into the general form (1.52) gives the action of =2 super Yang–Mills N theory. It can be viewed as a particular =1 supersymmetric gauge theory coupled N to chiral matter, namely it corresponds to the case of one single flavor in the adjoint representation of the gauge group. Supersymmetry does not allow a superpotential for the model. Integration over the θ variables gives the component expansion. The superfield Ψ is a singlet under the SU(2) R-symmetry group. The resulting component multiplet obtained eliminating the auxiliary fields in the Wess–Zumino gauge contains
a vector Aµ and a complex scalar φ that are SU(2) singlets and two Weyl spinors (ψ,λ) that transform in the 2 of SU(2). As already noticed, this field content is what would be obtained putting together a vector and a chiral =1 superfield both in the N adjoint of the gauge group. The most general =2 action that can be constructed in terms of the superfield N Ψ of equation (1.51) is 1 S = Im d4x d2θd2θ˜ (Ψ) , 4π F Z Z 1.6 Quantization of SUSY theories 29
where is a holomorphic function of Ψ called prepotential. F
=2 matter. N The =2 super Yang–Mills model can be generalized introducing the coupling to N matter described by hypermultiplets. Hypermultiplets are associated to superfields satisfying (1.49). A kinetic term for these superfields can be written in the form
4 αi j i βk k j SK = d x D Dα Φ †D DβΦ . Z Explicit dependence on the central charge does not allow a manifestly =2 formu- N lation without resorting to harmonic superspace. Anyway an on-shell action can be constructed in terms of = 1 superfields. Denoting the two chiral superfields in the N hypermultiplet by Q and Q˜ the action in =1 language is given by N
4 4 2V 2V 2V S = d x d θ Q†e Q + Q˜†e Q˜ + Φ†e Φ + Z Z 1 h i + Im τ d2θ W αW + d2θ √2Q˜ΦQ + m Q˜ Q + h.c. . (1.53) 4π α i i i Z Z The action (1.53) contains the ordinary kinetic terms, a mass term for chiral fields in the hypermultiplet and a superpotential that couples Q, Q˜ and Φ. Notice that no potential term coming from a self interaction of the fields Q and Q˜ is allowed by =2 supersymmetry. This follows from the observation that Q and Q˜ are assembled N into a superfield Φi transforming in the 2 of SU(2) R-symmetry and there exists no trilinear invariant in SU(2).
1.6 Quantization of supersymmetric theories
The functional approach to the quantization of field theories, based on the con- struction of a generating functional for the time-ordered products of fields, can be applied with no further difficulty to supersymmetric theories in their component field formulation. However, although this formulation allows to extract all the peculiar properties of supersymmetric theories at the quantum level, it makes rather obscure how such properties are a direct consequence of supersymmetry. On the contrary a quantization procedure can be carried out directly in superspace in a step by step manifestly supersymmetric fashion. This approach, that allows to construct directly 30 Chapter 1. Rigid 4d supersymmetry
Green functions for the superfields, has proved extremely powerful and it is perhaps the most useful application of superspace techniques. Super Feynman rules for the computation of Green functions of the superfields were first proposed in [14] and then reformulated in a more powerful way in [23]. This latter formulation will be reviewed here. Once the Green functions for the superfields are known contact with the “physical” component field formulation is established by further integrating the fermionic coordinates (θ, θ) to extract correlators for the various components. The discussion will focus on the quantization of =1 superfields; it is also possible N to generalize the method to -extended superspace, e.g. to harmonic superspace, N but the =1 formulation appears to be the most suitable for application to explicit N calculations, even in the extended case.
1.6.1 General formalism
In ordinary quantum field theories the path integral quantization is based on the
construction of a generating functional for the Green functions. Denoting by φi the generic fields of the model and by S[φ] the classical action the generating functional is defined as
Z[J]= N [ φ] eiS[φ]+i dxφi(x)Ji(x) , D Z R
where N is a normalization such that Z[0] = 1 and Ji(x) are external sources asso-
ciated to the fields φi(x). Green functions are obtained taking functional derivatives
with respect to the sources Ji. This approach has a natural extension that leads to a generating functional for the Green functions of the superfields in a supersymmetric theory. The general form of the action of a supersymmetric theory involves, in =1 N language, chiral and vector superfields and can be written
4 4 4 2 S = d xd θK(Φ, Φ†,V )+ d xd θ W[Φ] + h.c. . (1.54) Z Z To construct a generating functional classical sources for the different superfields must be introduced, that are in turn superfields. Since functional derivatives are to be taken with respect to the classical sources, such sources must be arbitrary. A problem arises with chiral superfields as sources coupled to them must themselves 1.6 Quantization of SUSY theories 31
be chiral. This difficulty can be solved by using a suitable definition of functional derivatives with respect to chiral superfields. For a generic superfield F (x, θ, θ) functional derivatives are defined according to the rule
δF (x′, θ′, θ′) = δ4(x x′)δ4(θ θ′) , δF (x, θ, θ) − − where the δ function of fermionic variables is to be interpreted as
2 2 δ (θ θ′)= δ (θ θ′)δ (θ θ′)=(θ θ′) (θ θ′) , 4 − 2 − 2 − − − so that
4 d θ δ (θ θ′)F (x, θ, θ)= F (x, θ′, θ′) 4 − Z and
δ 4 4 d x′d θ′ F (x′, θ′, θ′)G(x′, θ′, θ′)= G(x, θ, θ) . (1.55) δF (x, θ, θ) Z For chiral superfields Φ one defines in turn
δΦ(x′, θ′, θ′) 1 2 = D δ4(x x′)δ4(θ θ′) . δΦ(x, θ, θ) −4 − −
In this way the same result as in equation (1.55) is reproduced
δ 4 2 d x′d θ′ Φ(x′, θ′, θ′)Φ(˜ x′, θ′, θ′)= δΦ(x, θ, θ) Z 4 2 1 2 = d x′d θ′ D δ (x x′)δ (θ θ′)Φ(˜ x′, θ′, θ′)= −4 4 − 4 − Z 4 4 = d x′d θ′δ (x x′)δ (θ θ′)Φ(˜ x′, θ′, θ′)= Φ(˜ x, θ, θ) , 4 − 4 − Z 2 where in the last line it was used the property that d2θ and 1 D are equivalent under − 4 space-time integration, as can be proved by direct calculation by an integration by parts. Given these definitions of functional derivatives the generating functional can be written as
† † † iS[Φ,Φ ,V ]+i(Φ,J)+i(Φ ,J )+i(V,JV ) Z[J, J †,J ]= N [ Φ Φ† V ] e , (1.56) V D D D Z 32 Chapter 1. Rigid 4d supersymmetry
where S is the action (1.54), J (J †) is a chiral (antichiral) source, JV is real and
(Φ,J) = d4xd2θ Φ(x, θ, θ)J(x, θ, θ) , Z 4 2 (Φ†,J †) = d xd θ Φ†(x, θ, θ)J †(x, θ, θ) , Z 4 4 (V,JV ) = d xd θV (x, θ, θ)JV (x, θ, θ) . Z Green functions can then be calculated by the formula
(z ,...,z , z ,... ,z , z ,... ,z )= Gn 1 i i+1 j j+1 n = 0 T Φ(z ) . . . Φ(z )Φ†(z ) . . . Φ†(z )V (z ) ...V (z ) 0 = (1.57) h | { 1 i i+1 j j+1 n }| i δnZ[J, J ,J ] = ( i)n † V . − δJ(z1) ...δJ(zi)δJ †(zi+1) ...δJ †(zj)δJV (zj+1) ...δJV (zn) † J=J =JV =0
A generating functional for the connected Green functions can be obtained by taking the logarithm of Z
W [J, J †,J ]=( i) log Z[J, J †,J ] . V − V The quantum effective action Γ that acts as a generating funct ional for the one particle irreducible Green functions is obtained through the Legendre transform. Defining the “classical” fields δW δW δW Φ=˜ , Φ˜ † = , V˜ = , (1.58) δJ δJ † δJV Γ is given by ˜ ˜ ˜ Γ[Φ, Φ†, V ]= W [J, J †,JV ] (J, Φ)+(J †, Φ†)+(JV ,V ) ˜ ˜ † ˜ , − (Φ,Φ ,JV ) ˜ ˜ ˜ ˜ ˜ ˜ ˜ ˜ ˜ where on the right hand side J = J[Φ, Φ†, V ], J † = J †[Φ, Φ†, V ] and J V = JV [Φ, Φ†, V ] come from the inverting (1.58). The generating functional Z can be expressed in terms of the corresponding func-
tional in the free theory, Z0. The general action (1.54) can be divided up into a free part and an interacting part
S[Φ, Φ†,V ]= S0[Φ, Φ†,V ]+ Sint [Φ, Φ†,V ] ,
where S0 is quadratic in the fields and contains the kinetic and mass terms. Then one can write
δ δ δ iSint δJ , † , δJ Z[J, J †,JV ]= e δJ V Z0[J, J †,JV ] , (1.59) h i 1.6 Quantization of SUSY theories 33
where
† † † iS0[Φ,Φ ,V ]+i(Φ,J)+i(Φ ,J )+i(V,JV ) Z [J, J †J ]= N [ Φ Φ† V ] e 0 V D D D Z δ δ δ iSint [ δJ , † , δJ ] and e δJ V is to be understood as a power expansion. Since S0 is quadratic
Z0 is a Gaussian integral that can be explicitly computed. Writing
i 4 4 Φ Z0[J, J †,JV ] = exp d xd θ (Φ, Φ†) Φ + −2 M Φ† Z i 4 4 + d xd θV V + i(Φ,J)+ i(Φ†,J †)+ i(V,J ) (1.60) 2 MV V Z one gets
i 8 8 J(z′) Z0[J, J †,JV ] = exp d zd z′ (J(z),J †(z))∆Φ(z, z′) + 2 J †(z′) Z i 8 8 + d zd z′ J (z)∆ (z, z′)J (z′) , (1.61) 2 V V V Z where superfield propagators, ∆Φ and ∆V , have been introduced.
1.6.2 Superfield propagators
To compute the propagators the free action must be put in a canonical quadratic form. Since the computation leads to different kinds of difficulties for chiral and vector fields the two cases will be considered separately.
Chiral superfield propagator.
The general form of the free action is
4 4 2 1 2 1 S = d x d θ Φ†Φ + d θ m Φ Φ + d θ m Φ†Φ† = 0 i i 2 ij i j 2 ij i j Z Z Z Z 4 4 1 DD DD = d xd θ Φ†Φ m Φ Φ + Φ† Φ† , (1.62) i i − 8 ij i 2 j i 2 j Z 2 where the last line follows from the equivalence d2θ 1 D2, d2θ 1 D , valid ↔ − 4 ↔ − 4 under a space-time integration, and the action of the projection operators
2 1 D2D P = , P Φ† = Φ† , P Φ=0 1 16 2 1 1 2 1 D D2 P = , P Φ = Φ , P Φ† =0 . (1.63) 2 16 2 2 2 34 Chapter 1. Rigid 4d supersymmetry
Hence the action can be rewritten as
1 4 4 Φj S0 = d xd θ Φi, Φi† ij , (1.64) 2 M Φj† Z where 1 mij 4 2 DD δij ij = − . 1 mij M δij DD ! − 4 2 Considering for simplicity the case of one single flavor and taking into account the
source terms, the exponent of the Gaussian integral that yields Z0 turns out to be 1 D2 4 4 1 Φ 4 2 0 J i d xd θ Φ, Φ† + Φ, Φ† − 2 . 2 M Φ† 0 1 D J † Z " 4 2 ! # − A simple but lengthy calculation using D-algebra and properties of the projection
operators allows to write Z0 in the form of equation (1.61), where the chiral superfield propagator, known as Grisaru–Roˇcek–Siegel propagator [23], reads
m D2 1 4 2 1 ∆Φ(x, θ, θ; x′, θ′, θ′) = 2 δ8(z z′) (1.65) 2 m2 1 m D − − 4 2 ! and δ (z z′)= δ (x x′)δ (θ θ′). 8 − 4 − 4 −
Vector superfield propagator.
The free field action for a vector superfield is
1 1 α˙ S = d4x d2θ W αW + d2θ W W + d4θ m2V 2 . (1.66) 0 4 α 4 α˙ Z Z Z Z No massive vector field will enter the models considered in the following so m will be put to zero from the beginning. The kinetic operator in (1.66) has zero-modes so that just like in ordinary gauge theories a gauge fixing procedure `ala Faddeev–Popov is necessary. This requires the introduction of a gauge fixing term in (1.66) and a further integration over the Faddeev–Popov ghost fields will appear. The gauge-fixed action thus becomes
S = S0 + SGF + SFP = 1 1 α˙ = d4xd4θ W αW δ (θ)+ W W δ (θ) + 4 α 2 4 α˙ 2 Z ξ 2 + d4xd4θ (D V )(D2V ) + −8 Z 4 4 +i d xd θ (C′ + C′) V (C + C) + coth V (C C) , (1.67) L 2 L 2 − Z h i 1.6 Quantization of SUSY theories 35
where C, C′ are chiral superfields for the ghosts and no source term has been intro- duced for such fields. The gauge fixing term here introduced corresponds to a class of gauges that for the component field Aµ interpolate between the Lorentz gauge (ξ = ) and the Fermi–Feynman gauge (ξ = 1). ∞ The free propagators for the ghosts are exactly the same as for ordinary chi- ral superfields. Furthermore there is a non polynomial interaction with the vector superfield coming from the expansion of coth( ). LV/2 The vector superfield propagator is calculated from the quadratic part of the action
S = S0 + SGF = = d4xd4θ V [ 2P ξ(P + P )2] V = { − T − 1 2 } Z = d4xd4θV V . (1.68) MV Z Introducing the source term for V the generating functional becomes
i V V V +i(V,JV ) Z[J ]= [ V ] e M . V D Z R The Gaussian integral can now be performed yielding a result of the form (1.61), where the vector superfield propagator is 1 α ∆ (x, θ, θ; x′, θ′, θ′) = P (P + P ) δ (z z′)= V −2 T − 2 1 2 8 − 1 = [1+(α 1)(P + P )] δ (z z′) , (1.69) −2 − 1 2 8 − 1 with α = ξ .
1.6.3 Feynman rules
Once the free field propagators are known the perturbative expansion for the Green functions of the superfields can be derived from the generating functional in the form (1.59) that yields
(z ,...,z , z ,...,z , z ,... ,z )= Gn 1 i i+1 j j+1 n = 0 T Φ(z ) . . . Φ(z )Φ†(z ) . . . Φ†(z )V (z ) ...V (z ) 0 = h | { 1 i i+1 j j+1 n }| i δ δ δ δ δ δ = ...... δJ(z1) δJ(zi) δJ †(zi+1) δJ †(zj) δJV (zj+1) δJV (zn) · k ∞ ( i)k δ δ δ − Sint , , Z0[J, J †,JV ] . (1.70) · k! δJ δJ † δJV k=0 † J=J =JV =0 X
36 Chapter 1. Rigid 4d supersymmetry
A super Feynman graph technique can be developed which is a straightforward gen- eralization of what is done in ordinary field theories. Super Feynman rules for the evaluation of one particle irreducible Green functions can be derived given the above generating functional. The rules can be summarized as follows.
To each external line is associated the corresponding superfield. • At each vertex there is an integration over the whole superspace d4xd4θ; factors, • as for example powers of the coupling constant, to be associated to the vertices, are read directly from the action.
2 A factor of 1 D or 1 D2 acting on the internal chiral or antichiral lines • − 4 − 4 respectively must be included at each vertex. Such factors come from the definition of the functional derivatives. At purely chiral (antichiral) vertices one 2 factor of 1 D ( 1 D2) must be suppressed, because it is used to reconstruct − 4 − 4 the measure over the whole superspace.
For internal lines the propagators previously derived are used. • The specific form of the interaction at each vertex as well as the combinatoric • analysis are worked out exactly like in ordinary perturbation theory.
The expressions for the one particle irreducible Green functions in momentum • space are obtained by Fourier transforming in the x variables, but not in the fermionic θ coordinates.
1.7 General properties of supersymmetric theories
Supersymmetry has far reaching consequences in the quantum theory both at the perturbative and at the non-perturbative level. Some general results will be briefly recorded in this section. More can be found in the books and reviews quoted in the references. Some dramatic effects on the divergences encountered in perturbation theory can be easily derived from the Feynman rules. Explicit calculation shows that in Feynman diagrams four covariant derivatives are associated at each vertex, with the exception of vertices involving external (anti) chiral lines that have two derivatives less. More- 2 D2 D over ΦΦ (Φ†Φ†) propagators with momentum p have an additional factor of p2 ( p2 ). 1.7 General properties 37
Simple dimensional analysis allows to compute the superficial degree of divergence ds of a diagram which is
d =4L 2P +2V C E 2L , s − − − − where L is the number of loops, P the number of internal propagators, V the number of vertices, C the number of ΦΦ or Φ†Φ† propagators and E the number of external chiral or antichiral lines. Use of the topological constraint L P + V = 1 yields [24] − d =2 C E . s − − For graphs with only external V lines gauge invariance requires the presence of four D or D factors on the external lines so that the result is
d = C s − 2 For diagrams with only lines of a given chirality there are further factors of D2 or D associated to the external lines so that one obtains
d =1 C n , s − − where n is the number of external chiral or antichiral lines. This calculation allows to determine all the potentially divergent diagrams. The only possible divergences are logarithmic and can arise in diagrams with only external V lines with C =0(e.g. corrections to the vector superfield propagator) and in
Φ†Φ propagators with C = 0. This shows that in supersymmetric models only wave function renormalizations may be required; there is no direct mass or coupling 2 constant renormalization, i.e. with W = mφ one has ZW = 1, but Zφ = 1 implies 1/2 6 Z = Z− = 1. m φ 6 This and other results can be obtained from the general non renormalization theorem for =1 theories [25]: Any perturbative quantum contribution to the effective N action Γ can be expressed as an integral over the whole superspace and is local in the θ variables, namely
Γ = d4x ...d4x d4θG(x ,...,x ) 1 n 1 n · n X Z f [Φ, Φ†,V,D Φ, D Φ†,D V, D V,... ] . (1.71) · n α α˙ α α˙ The theorem can be easily proved using Feynman rules and successively integrating by parts the θ variables exploiting the properties of Grassmannian integration. Explicit 38 Chapter 1. Rigid 4d supersymmetry
calculations in the =4 supersymmetric Yang–Mills theory showing the mechanism N leading to this result will be presented in chapter 3, so the proof of the above statement will not be discussed here. Notice that the non renormalization theorem does not forbid in principle quantum corrections to the superpotential. It is easy to see that a contribution of the form of (1.71) can produce a correction to the superpotential [26, 27]. As an example consider a term of the form
2 D2 D D2 d4xd2θd2θ Φn = d4xd2θ Φn = d4xd2θΦn , 162 162 Z Z Z 2 D D2 where Φ is chiral. The right hand side, which follows since P1 = 162 is the projector on chiral superfields, has the form of a correction to the superpotential. This example shows that in principle an arbitrary correction to the superpotential can be generated perturbatively. This effect is related to an infrared singular behaviour that can be present only in theories containing massless fields. The example shows that the correction, which is non-local when written as an integral over the whole superspace, can be rewritten as a local term integrated over a subspace. The same construction could not be repeated if the theory does not contain massless particles, since in that 1 case one would obtain a factor of 2+m2 , which cannot be re-expressed as a subintegral. In [27] it was shown that such a correction as the one considered above actually occurs at the two-loop level in the Wess–Zumino model. In a =1 super Yang–Mills theory N coupled to the Wess–Zumino model the correction appears at one loop [28]. An alternative approach to the study of this kind of effects has been given in [29, 30]. Other non renormalization theorems hold for theories of extended supersymmetry. In particular considerations based on the so called anomalies argument allow to prove the finiteness of =2 theories beyond one loop. More precisely it can be proved that N the only perturbative contributions to the β function of extended supersymmetric theories can come from one loop diagrams. This result allows to single out a class of finite supersymmetric models and among them the =4 super Yang–Mills theory N that will be studied in detail in the following chapters. Supersymmetry puts severe restrictions on the non-perturbative properties as well. A review of the general results on the non-perturbative dynamics of supersymmetric theories is provided by [31, 32]. Many astonishing results in this context can be traced back to the properties of holomorphy of supersymmetric theories. In supersymmetric models for example the superpotential is a holomorphic function of the chiral super- 1.7 General properties 39
fields. Analogously, since the parameters of the theories (such as masses and coupling constants) can be viewed as expectation values of auxiliary fields, various quantities depend analytically on such parameters. Using holomorphy properties many exact result have been derived. In [33] an expression for the β function of supersymmetric gauge theories was proposed that is exact at the perturbative as well as at the non- perturbative level for a class of models. Also on holomorphy relies the construction of the exact solution for the low energy dynamics of a class of =2 theories proposed N by Seiberg and Witten [34]. Chapter 2
=4 supersymmetric Yang–Mills N theory
Four dimensional =4 supersymmetric Yang–Mills theory is a very special quantum N field theory. It possesses several peculiar properties that will be reviewed in this chapter. The general properties of the model are discussed here, original results in the study of =4 Yang–Mills will be presented in the following chapters. N The action of the theory was given for the first time in [35, 36] within the frame- work of string theory toroidal compactifications. The theory has the maximal amount of supersymmetry for a rigid supersymmetric theory in four dimensions, namely six- teen real supercharges; a larger number of supercharges would require fields of spin larger than one and therefore the inclusion of gravity. Historically the interest in this model was raised by its property of finiteness; the β function of Gell-Mann and Low has been proved to be vanishing in perturbation theory and the same is supposed to be true at the non-perturbative level. As a con- sequence the superconformal invariance displayed by the classical theory is believed to be preserved after quantization. Moreover =4 super Yang–Mills theory possesses exact electric-magnetic duality. N It contains beyond the elementary fields an infinite set of dyonic states which are believed to realize exactly the duality symmetry proposed by Montonen and Olive in [37]. More recently there has been a renewal of interest in =4 Yang–Mills theory N 40 2.1 Formulations of =4 SYM 41 N following the proposal made by Maldacena in [38] of a new duality relating type IIB supergravity in d + 1 dimensional anti-de Sitter space and d dimensional (su- per)conformal theories. This subject will be discussed in detail in chapter 5. The present chapter is organized as follows. In section 2.1 the various formulations of the model, which will be employed at different stages in the following, are reviewed. Section 2.2 presents the classical symmetries and conserved currents of the theory. The properties of finiteness are discussed in section 2.3. The final section is then devoted to electric-magnetic duality in =4 supersymmetric Yang–Mills theory. The N general aspects of electric-magnetic duality are reported separately in appendix B.
2.1 Diverse formulations of =4 supersymmetric N Yang–Mills theory
=4 supersymmetric Yang–Mills theory in four dimensions was obtained for the first N time in [35, 36] by applying the method of dimensional reduction to =1 super Yang– N Mills in ten dimensions. The latter is the low energy effective theory coming from type I superstring theory and describes a = 1 vector multiplet in ten dimensions N consisting of one real vector and one Majorana–Weyl spinor. The action is 1 1 i S = d10x tr F F ΓΛ + λΓΛD λ , (2.1) −4 ΓΛ 2 Λ Z where F = ∂ A ∂ A + ig[A , A ] and λ satisfies the Majorana and Weyl ΓΛ Γ Λ − Λ Γ Γ Λ conditions
T λ = C λ , λ = Γ λ , (10) ± 5 where C(10) is the charge conjugation operator. The action (2.1) is invariant under the supersymmetry transformations
δηAΛ = iηΓΛλ ΓΛ δηλ = ΣΓΛF η . (2.2)
In the previous equations capital Greek letters Γ and Σ denote ten dimensional ma- trices. Dimensional reduction ´ala Kaluza–Klein on a six dimensional torus T 6 leads to a multiplet of fields in four dimensions possessing an additional SU(4) SO(6) global ∼ 1Capital Greek indices refer to ten dimensional space-time. 42 Chapter 2. =4 super Yang–Mills theory N
symmetry, which is a direct consequence of the ten dimensional Lorentz invariance. More precisely it is a consequence of the fact that the torus T 6 has a trivial holon- omy group. In general in the dimensional reduction on a Riemannian manifold of dimension k the spin connection is a SO(k) gauge field and consequently the spinors transform, upon parallel transport around a closed contractible curve, under a sub- group of SO(k), which is the holonomy group. In the case at hand of T 6 every spinor is covariantly constant, so that the whole SO(6) SU(4) becomes a global symmetry ∼ of the model. From the point of view of the =4 theory this SU(4) global symmetry N is identified with the R-symmetry group of the =4 supersymmetry algebra, as will N be discussed in the next section. In the bosonic sector the compactification gives rise to a four dimensional real vector and to six real scalars from the internal components of the ten dimensional vector. More precisely one defines
a a Aµ = Aµ µ =0, 1, 2, 3 ϕa = 1 (Aa + iAa ) B =1, 2, 3 B4 √2 B+3 B+6 a AB 1 ABC4 a a AB ∗ ϕ = 2 ε ϕC4 = ϕ A,B,C =1, 2, 3 . (2.3)
In the fermionic sector a suitable choice of the ten-dimensional Γ matrices allows to write, in Majorana notation, the 32-component Majorana–Weyl spinor λ as Lχ1 . . 4 Lχ A T λ = , χ˜A = C(χ ) , Rχ1 . . Rχe 4 where L and R denote the left and right chirality projection operators and C is e the four-dimensional charge conjugation operator related to the ten-dimensional one 0 1I4 by C(10) = C . In four dimensions one obtains four Majorana (or ⊗ 1I4 0 equivalently Weyl) spinors
LχA λA = A =1, 2, 3, 4 . (2.4) RχA ! At the end of the day the result of the compactification is the = 4 multiplet e N a AB a which contains six real scalars ϕ satisfying (2.3), one real vector Aµ and four 2.1 Formulations of =4 SYM 43 N
aA Weyl spinors λα , all in the adjoint representation of the gauge group. The construc- tion here described from the ten dimensional theory allows to derive immediately the transformation of the fields under the SU(4) global symmetry. Because of the constraint (2.3) the scalars ϕa AB are in the second rank complex self dual 6 of SU(4); aA a a the spinors λ are in the 4 while the λA transform in the 4; the gauge field Aµ is a singlet. In the following a different notation for the scalar fields will also be used, namely one can define 1 ϕi = ti ϕAB , i =1, 2,..., 6 , 2 AB AB where (ti) are Clebsch–Gordan coefficients that couple two 4’s to a 6 (these are µ six-dimensional generalizations of the four-dimensional σ αα˙ matrices). In Weyl notation the action of the four dimensional =4 theory turns out to be N 1 ←→ α˙ 1 S = d4x tr (D ϕAB)(Dµϕ ) i(λαA D/ λ ) F F µν + µ AB − 2 αα˙ A − 4 µν Z n α˙ gλαA[λ B, ϕ ] gλ [λ ,ϕAB]+2g2[ϕAB,ϕCD][ϕ , ϕ ] , (2.5) − α AB − αA˙ B AB CD ab ab oabc where tr denotes a trace over the colour indices and Dµ = δ ∂µ + igf Aµc is the covariant derivative in the adjoint representation. The action (2.5) is invariant under the supersymmetry transformations
1 1 α˙ δϕAB = (λαAη B λαBη A)+ εABCDη λ 2 α − α 2 α˙ C D A 1 µν β A AB α˙ CA B δλ = F − σ η +4i(/D ϕ )η 8g[ϕ ,ϕ ]η α −2 µν α β αα˙ B − BC α α˙ δAµ = iλαAσµ ηα˙ iηαAσµ λ , (2.6) − αα˙ A − αα˙ A which can be easily obtained from the ten dimensional transformations (2.2) using (2.3) and (2.4). A manifestly off-shell formulation of =4 supersymmetric Yang–Mills theory N would require the introduction of an infinite set of auxiliary fields as a consequence of the presence of bosonic coordinates associated to the twelve central charges in the =4 superspace, as noticed in [39, 40]. A superfield formulation can be constructed N using analytic superspace, however such an approach is rather involved and cannot be applied to explicit calculations because it is intrinsically on-shell. An on-shell AB AB A superfield description can be given in terms of a superfield W = W (x, θα , θαA˙ ) satisfying [41] the reality condition 1 W = ε W CD , (2.7) AB 2 ABCD 44 Chapter 2. =4 super Yang–Mills theory N
together with the constraint,
AW BC = [AW BC] , (2.8) Dα Dα
where is the super-covariant derivative. The fields entering the action (2.5) are Dα combined into the lowest components of W AB. In the next chapter perturbative calculations in =4 Yang–Mills will be carried N out using a formulation of the model in terms of = 1 superfields. The field content N of the theory can be obtained by taking one =1 vector superfield V and three N =1 chiral superfields ΦI all in the adjoint representation of the gauge group. In N this approach the six real scalars are combined into three complex fields that are the scalar components of the chiral superfields ΦI , three of the Weyl fermions are the spinors of ΦI and the fourth fermion is the gaugino that together with the real vector constitute the vector superfield V . As a result only a SU(3) U(1) subgroup × of the original SU(4) symmetry is manifest in this formulation. More precisely the representations of SU(4) decompose according to 6 3+3, 4 3+1, so that the → → I chiral superfields Φ transform in the 3 of SU(3), the antichiral ΦI† in the 3 and the vector V is a singlet under SU(3). The action in the =1 superfield formulation reads N
1 4 4 gV gV I 1 2 1 α S = tr d x d θ e− ΦI† e Φ + d θ W Wα + h.c. + dr 4g2 4 Z Z √2 2 I J K 2 IJK + ig d θε Φ [Φ , Φ ]+ d θε Φ† [Φ† , Φ† ] , (2.9) 3! IJK I J K Z Z #)
where dr denotes the Dynkin index of the representation to which the fields belong. To recover the correct dependance on the coupling constant g one must substitute v 2gV in the second term in (2.9). → The action (2.9) is non-polynomial as is always the case in supersymmetric gauge theories in =1 superspace. In particular the scalar potential of (2.5) comes from N the superpotential term
d4x d2θ W(x, θ, θ) + h.c. = Z Z √2 = d4x d2θ i ε f abc(ΦI ΦJ ΦK)(x, θ, θ) + h.c. − 3! IJK a b c Z (Z " # ) 2.1 Formulations of =4 SYM 45 N
By power expanding eV one obtains
S = d4xd4θ V a(x, θ, θ) [ 2P ξ(P + P )2] V (x, θ, θ)+ − T − 1 2 a Z a I a b Ic +Φ†I (x, θ, θ)Φa(x, θ, θ)+ igfabcΦ†I (x, θ, θ)V (x, θ, θ)Φ (x, θ, θ)+ 1 2 e a b c Id g f f Φ† (x, θ, θ)V (x, θ, θ)V (x, θ, θ)Φ (x, θ, θ)+ . . . −2 ab ecd I i 2 gf D DαV a(x, θ, θ) V b(x, θ, θ) D V c(x, θ, θ) + −4 abc α 1 h i 2 g2f ef V a(x, θ, θ) DαV b(x, θ, θ) D V c(x, θ, θ) D V d(x, θ, θ) + −8 ab ecd α √2 h i + . . . gf abc ε ΦI (x, θ, θ)ΦJ (x, θ, θ)ΦK (x, θ, θ)δ(θ)+ − 3! IJK a b c IJK a + ε Φ†Ia(x, θ, θ)Φ†Jb(x, θ, θ)Φ†Kc(x, θ, θ)δ(θ) + C′a(x, θ, θ)C (x, θ, θ)+
a i a a C′a(x, θ, θ)C (x, θ, θ) + gfabc C′ (x, θ, θ)+ C′ (x, θ, θ) − 2√2 · b c c 1 2 e a a V (x, θ, θ) C (x, θ, θ)+ C (x, θ, θ) g f f C′ (x, θ, θ)+ C′ (x, θ, θ) · − 8 ab ecd · d V b(x, θ, θ)V c(x, θ, θ) Cd(x, θ, θ)+ C (x, θ, θ) + . . . , (2.10) · which of course can be made polynomial by choosing the Wess–Zumino gauge which implies V 3 = 0. In (2.10) the gauge fixing term as well as the corresponding ghost fields have been included following the prescription of (1.67) of chapter 1 and the expansion has been truncated to the terms that will be relevant in future calculations. Integrating over the θ variables and eliminating the auxiliary fields through the equations of motion results in a different component field formulation, which again has a manifest SU(3) U(1) global symmetry. Such a formulation will be employed × in the next chapter in the discussion of subtleties related to the problem of gauge fixing in this model. Choosing the Wess–Zumino gauge one obtains
4 1 a a µν µ I i a a i a aI S = d x F F D ϕ† D ϕ + λ Dλ/ + ψ Dψ/ + −4 µν − µ I 2 2 I Z a a i√2 a +ig√2f λ ϕb ψcI ψ ϕbI λc f εI ψ ϕbJ ψcK + (2.11) abc I − I − 2 abc JK I IJ a b cK 1 2 a b 2 1 2 a LMK bI cJ d e ε ψ ϕ† ψ g (f ϕ† ϕ ) + g f f ε ε ϕ ϕ ϕ† ϕ† . − K I J − 2 abc I I 2 abc de IJK L M This formulation can be easily seen to be perfectly equivalent to (2.5) apart from the the manifest R-symmetry, that is only SU(3) U(1). × 46 Chapter 2. =4 super Yang–Mills theory N
It is also useful to give a third component field formulation of =4 supersym- N metric Yang–Mills theory in terms of = 2 multiplets. This approach will actually N be used in instanton calculations in chapter 4. The =4 super Yang–Mills multiplet N decomposes in terms of =2 multiplets into a vector plus a hypermultiplet. In this N description the subgroup of the original SU(4) R-symmetry that is realized explicitly is SU(2) SU(2) U(1), where the subscripts and refer to the vector and hy- V × H× V H permultiplet respectively: fields in the two multiplets are charged only with respect to one of the two SU(2) factors. The representations of SU(2) SU(2) U(1) will V × H× be denoted by (r , r )q with the subscript q referring to the U(1) charge and the V H following notation for the component fields will be used
λu (2, 1) , ϕ (1, 1) , A (1, 1) V → ∈ +1 ∈ +2 µ ∈ 0 u˙ ψ (1, 2) 1, quu˙ (2, 2)0 , (2.12) H → ∈ − ∈ where u, u˙ =1, 2. In particular the form of the Yukawa couplings in this formulation will be of relevance in the following calculations. In =2 notation the Yukawa N couplings read
( =2) √2 a bα u S cu˙ b Suu˙ cα˙ N = gf q λ σ ψ + ψ σ λ + LY 2 abc S uu˙ α α˙ u˙ u n a uv b cα˙ bα u˙ c v˙ + ϕ ε λαu˙ λv + εu˙v˙ ψ ψα +
a bα u c v u˙v˙ b cα˙ + ϕ εuvλ λα + ε ψα˙ u˙ ψv˙ , (2.13) o In conclusion of this review of the diverse formulations of = 4 super Yang– N Mills it is worth recalling that a truly off-shell analysis of the theory, in components and in the Wess–Zumino gauge, was given in [42] within the framework of algebraic renormalization. The problems related to the necessity of including an infinite set of auxiliary fields are dealt with using the method of Batalin and Vilkoviski. This approach, though technically rather involved, allows to prove that all the classical symmetries of the theory survive quantization.
2.2 Classical symmetries and conserved currents
The field content of =4 supersymmetric Yang–Mills theory described in the pre- N ceding section is unique apart from the choice of the gauge group G. In the Abelian case the theory is free as follows immediately by observing that all the fields belong 2.2 Symmetries and currents 47
to the adjoint representation of the gauge group, so that all the covariant derivatives reduce to ordinary ones if G is Abelian, and all the potential terms are proportional to the structure constants fabc. In the non Abelian case the theory has a moduli space of vacua parameterized by the vacuum expectation values (vev’s) of the six real scalars, corresponding to a Coulomb phase. The manifold of vacua is determined by the condition of vanishing scalar potential (F -flatness plus D-flatness) which reads
tr [ϕAB,ϕCD][ϕ , ϕ ] =0 tr [ϕi,ϕj][ϕ ,ϕ ] =0 . AB CD ⇐⇒ i j By writing ϕi explicitly in terms of generators of the adjoint representation one obtains
1,6 dim G 2 2 a b 2 ϕiaϕjb tr [T , T ] =0 , (2.14) i,j a=1 X X so that for a rank r gauge group G at most r components ϕia for each field can have non vanishing vacuum expectation value. As a result the space of vacua is parameterized by 6r real coordinates. The moduli space is
= R6r/ , (2.15) M Sr where is the group of permutations of r elements, because configurations in which Sr two or more components of ϕi are exchanged are physically indistinguishable. In con- clusion the manifold isa6r-dimensional real space with orbifold type singularities. M In a generic point of the Coulomb phase with nonvanishing vev’s for the scalar fields the gauge group G is broken to U(1)r, so that the theory is again Abelian and thus free. On the contrary the origin of the moduli space, where the vev’s of all the scalars vanish, corresponds to a phase that is expected to describe a highly non trivial superconformal field theory. Note that the superconformal phase corresponds to an orbifold singularity of the manifold . In this phase the classical action (2.5) is M actually invariant under a larger group of global transformations than that generated by the super Poincar´ealgebra (1.3), namely the =4 superconformal group. N The -extended superconformal algebra in four dimensions is an extension of N i (1.3) which includes additional generators D, Kµ, Sα and Sαi˙ corresponding respec- tively to dilatations, special conformal transformations and the associated special supersymmetry transformations. The complete algebra further includes the follow- 48 Chapter 2. =4 super Yang–Mills theory N
ing (anti)commutation relations, supplementing (1.3)
[M ,K ]= η K η K µν λ νλ µ − µλ ν [D,P ]= P [D,K ]= K µ − µ µ µ [P ,K ]= 2M +2η D [K ,K ]=0 µ ν − µν µν µ ν ˙ µν µν β i µν α˙ βi [Sαi, M ]=(σ )α Sβi [Sα˙ , M ]=(σ) β˙ S
j S , S =2σµ K δj { αi α˙ } αα˙ µ i i j S ,S =0 S , S ˙ =0 { αi βj} { α˙ β} 1 1 [Qi ,D]= Qi [Q ,D]= Q α 2 α αi˙ 2 αi˙ 1 i 1 i [S ,D]= S [S ,D]= S αi 2 αi α˙ 2 α˙ − (2.16) αi˙ α˙ [Qi ,Kµ]= iσµ S [Q ,Kµ]= iσµ αα˙ S α αα˙ i − αi µ αi˙ µ [Sαi,K ]=0 [S ,K ]=0
α˙ αi˙ [S ,P µ]= iσµ Q [S ,P µ]= iσµ αα˙ Qi αi − αα˙ i α a aj a i ai j [S , T ]= B S [T , S ]= B† S αi − i αj α˙ − j α˙ Qi ,S =2ε δi D i(σµν ) γε M δi 4iε δi A + Bai T { α βj} αβ j − α γβ µν j − αβ j j a j j µν γ˙ j j aj Q , S ˙ =2ε ˙ δ D iε (σ ) ˙ M δ +4iε ˙ δ A + B† T { αi˙ β} α˙ β i − α˙ γ˙ β µν i α˙ β i i a 4 4 [Qi , A]= i −N Qi [Q , A]= i −N Q α − 4 α αi˙ 4 αi˙ N N 4 i 4 i [S , A]= i −N S [S , A]= i −N S . αi 4 αi α˙ − 4 α˙ N N The operator A in these relations is the generator of the U(1) factor of the group U( )=SU( ) U(1) of the automorphisms of the algebra. The =4 case is singular N N × N since it implies
i [Qα, A] = [Qαi˙ , A]=0 .
The chiral rotation operator A must therefore have the same action on all the states in the multiplet. This implies in particular that it must act in the same way on states of opposite helicities 1 and 1 , which is possible only if A is actually zero. − 2 2 2.2 Symmetries and currents 49
For this reason the R-symmetry group in the = 4 case is SU(4) and not U(4) as N would be expected from general arguments. As already noticed in section 2.1 a SU(4) SO(6) global symmetry, that is identi- ∼ fied with the R-symmetry, naturally arises in the compactification on T 6 that leads to =4 super Yang–Mills in D = 4 from the ten dimensional =1 theory. In this N N context the reduction of the global symmetry from U(4) to SU(4) is understood as a consequence of the lack of a real six dimensional representation in U(4). More αA B precisely it follows from the form of the Yukawa coupling λ [λα , ϕAB] in the ac- tion (2.5). In fact a transformation, under the U(1) factor contained in U(4), of the iα AB 2iα AB type λ e λ for the spinor would require ϕ e− ϕ for the scalars which is → → prohibited by the reality condition (2.3). The Noether currents associated with the superconformal transformations to- gether with those corresponding to chiral SU(4) R-transformations are components of a unique on-shell multiplet [43]. The complete multiplet of currents was given for the Abelian case in [44]. The conserved currents are
←→ µν 1 µν 2 µ νρ 1 αA (µ ν) α˙ = [δ (F − ) 4F − F − + h.c.] λ σ ∂ λ T 2 ρσ − ρ − 2 αα˙ A + δµν (∂ ϕ )(∂ρϕAB) 2(∂µϕ )(∂ν ϕAB) ρ AB − AB 1 (δµν2 ∂µ∂ν )(ϕ ϕAB) − 3 − AB ←→ µ κν µ α˙ µ B 4 µν β B Σ = σ F − σ λ +2iϕ ∂ λ + iσ ∂ (ϕ λ ) αA − κν αα˙ A AB α 3 α ν AB β ←→ 1 α˙ µ B = ϕ ∂µϕCB + λ σµαα˙ λ B δ BλαC σµ λ . (2.17) J A AC αA˙ α − 4 A αα˙ C The remaining components of the supermultiplet are obtained by supersymmetry using the equations of motion and read
2 = (F − ) C µν A µν β A Λˆ = σ F − λ α − α µν β AB = λαAλ B E α AB αA β B AB = λ σ λ +2iϕ F − (2.18) Bµν µνα β µν 1 χˆC = ε (ϕDEλ C + ϕCEλ D) α AB 2 ABDE α α 1 AB = ϕABϕ δA δB ϕEF ϕ . Q CD CD − 12 [C D] EF In (2.17) is the (improved) energy momentum tensor, Σµ are the spin 1 su- Tµν αA 2 persymmetry currents and µB are the SU(4) currents. is a singlet under SU(4) JA Tµν 50 Chapter 2. =4 super Yang–Mills theory N
whereas Σµ transforms in the 4 and µB in the 4 4. The multiplet further includes αA JA × (AB) ij 1 C ˆ A three scalars , and , two fermionic spin 2 componentsχ ˆα AB and Λα and C E Q[AB] one antisymmetric tensor µν . Their transformation under SU(4) is as follows: the B scalars , (AB) and ij are respectively in the 1, 10 and 20, the two fermionsχ ˆC C E Q α AB A [AB] and Λˆ belong to the 4 6 and 4 respectively and µν is in the 6. α × B All of these fields can be obtained as lowest components of an on-shell superfield ij which is a bilinear in W i 1 ti W AB W(2) ≡ 2 AB δij ij = tr W iW j W W k . W(2) − 6 k In the non Abelian case in (2.17) and (2.18) there is a trace over the colour indices and additional terms are present. Such terms comprise beyond those required for the covariantization of all the derivatives also ‘potential’ terms. For example the complete non Abelian expression for (AB) is E (AB) = tr λαAλ B + g t(AB)+ tr ϕiϕjϕk , (2.19) E α [ijk] (AB)+ AB where t[ijk] is the completely antisymmetrized product of three ti matrices.
2.3 Finiteness of =4 supersymmetric Yang–Mills N theory
The first argument suggesting the possibility that the =4 supersymmetric Yang– N Mills theory could be a finite theory at the quantum level, free of ultraviolet as well as infrared divergences, was given in [19]. In that paper within the study of the general problem of the coupling of scalar (chiral) multiplets to vector multiplets it was observed that a number of three chiral multiplets in a supersymmetric Yang– Mills theory would lead to a Gell-Mann and Low β function vanishing in the one-loop approximation. In fact for a gauge group G=SU(N) the one loop β function is given by g2 β = (3 n)N , −16π2 − where n is the number of chiral multiplets coupled to the vector multiplet in the =1 N language. Successively this result was improved in [45], where the same was shown to hold at the two loop level by an explicit computation in the component field formulation. 2.3 Finiteness of =4 SYM 51 N
Further improvement of this result beyond two loops proves extremely complicated using ordinary techniques based on Feynman rules for the component theory. There exist a computer calculation [46] of the three-loop β function performed by using ordinary Feynman rules which gives a vanishing result as well. Superspace techniques allow to derive this result in a much simpler way as was shown in [23, 47, 48]. The basic ingredient is the observation that a vertex coupling three chiral superfields is finite as a direct consequence of dimensional analysis. By applying the rules of section 1.7 the superficial degree of divergence of a Green func- tion with three external (anti) chiral superfields is negative, D = 2 3 = 1, i.e. − − the diagram is finite. This means that the β function is completely determined by the wave function renormalization constant calculated from the divergent part of the propagator of, say, the chiral superfield. The equation for the β function [48] is
3 ∂Z(1) β(g)= g2 g , 2 ∂g
(1) where Zg is the coefficient of the simple pole term in the Fourier transform of the chiral superfield propagator. Calculations presented in the papers of references [47, 48] show that the propagator of the chiral superfield is finite, in the supersymmetric generalization of the Fermi–Feynman gauge (α=1 in the notation of section 1.6), up to three loops, implying the vanishing of the β function at the same order in perturbation theory. The lack of manifest =2 supersymmetry in the calculations described up to now N prevents from using the non renormalization theorem for =2 in order to extend the N result to every order. A formulation based on unconstrained = 2 superfields (which N is basically a version of harmonic superspace) was developed in [49]; such an approach allows a description of =4 super Yang–Mills with manifest =2 supersymmetry. N N The finiteness of the theory at all orders in perturbation theory is then a consequence of the =2 non renormalization theorem once the absence of divergences is proved N at one loop. Another proof of the finiteness of the theory was given in [50] using a light-cone superspace formulation. The advantage of this formulation relies on the fact that only physical fields are involved so that no auxiliary fields are necessary, however the lack of Lorentz invariance introduces other difficulties and makes the interpretation of the results less transparent. A different approach based on anomaly arguments has been presented in [51]. 52 Chapter 2. =4 super Yang–Mills theory N
The vanishing of the β function is obtained as a consequence the superconformal invariance since (see [52])
β(g) F a F µν = µ , g µν a Tµ
so that invariance under dilatations, implying the vanishing of µ, immediately gives Tµ β(g) = 0. The problem is therefore the proof of superconformal invariance, which is obtained as a consequence of the SU(4) symmetry being preserved at the quantum level. As already remarked a different demonstration of superconformal symmetry was given in [42] using somewhat more general arguments. All of the results here discussed either apply to gauge invariant quantities or are obtained avoiding possible subtleties related to the choice of the gauge. As will be discussed in the next chapter subtleties are actually present in the calculation of non gauge invariant quantities such as Green functions both in the component field formulation and in the description in terms of =1 superfields. N
2.4 S-duality
During the last few years the concept of duality has been the basic ingredient for striking results in the analysis of the non-perturbative behaviour of supersymmetric gauge theories as well as of superstring theories. The fundamental idea of dual- ity is that of a transformation relating two different and somehow complementary descriptions of a physical system. The rˆole that duality plays in the study of the non-perturbative properties of a physical model relies on the possibility of construct- ing a transformation that maps the non-perturbative regime of the system into the perturbative regime of the ‘dual’ system. Examples of this kind of duality have been known for a long time; in particular the duality shown by Kramers and Wannier [53] in the case of the 2-D Ising model and that established between the Sine-Gordon and the Thirring models [54, 55] appear very representative.
The two dimensional Ising model is defined by a set of spin variables σi taking values 1 and living on a two dimensional square lattice. The interaction is restricted ± to nearest neighbours and the partition function reads
Z(K)= exp K σ σ , (2.20) i j σ (i,j) X{ } X 2.4 S-duality 53
where K is related to the temperature T and the ferromagnetic coupling constant J through K = J/k T (k being Boltzmann’s constant). In (2.20) the sum on σ is B B { } over all spin configurations and the sum on (i, j) is over all pairs of nearest neighbours. The model was explicitly solved by Onsager and exhibits a single phase transition to a ferromagnetic state at a critical temperature Tc. Kramers and Wannier computed the critical temperature by using duality arguments, before the exact solution of the model was found. The crucial point is that the theory can be reformulated in a different (“dual”) way by writing the partition function as a sum over the elementary plaquettes of a dual lattice whose sites are the centers of the plaquettes of the original one, with a new coupling K∗. The two descriptions are then equivalent, namely
Z∗(K∗)= Z(K), if the couplings are related by 1 sinh 2K∗ = . sinh 2K Notice that this implies that the high temperature (or weak coupling) regime, K 1, ≪ of the original description is mapped by the duality transformation to the low temper- ature (or strong coupling) regime, K∗ 1, of the dual system. The computation of ≫ the critical temperature is based on the hypothesis that the system has a single tran- sition point. The critical point must then be the self-dual point of the transformation,
K = K∗, which gives sinh[2J/kBTc]=1. A second interesting example of duality occurs in two dimensional relativistic field theory and relates the Sine-Gordon and the Thirring models. The Sine-Gordon model is defined by the 2-D action 1 α S = d2x ∂ φ∂µφ + (cos βφ 1) , (2.21) SG 2 µ β2 − Z where φ is a scalar field and the parameter β plays the rˆole of a coupling constant as can be established by power expanding the potential term. The spectrum of the theory contains ‘meson’ excitations of mass Mm = √α and solitonic states with 8√α mass Ms = β2 . Therefore the solitonic degrees of freedom have large mass in the weak coupling regime. It can be shown that the Sine-Gordon model is completely equivalent to the Thirring model which describes a system of interacting fermions. The action of the latter is g S = d2x ψiγ ∂µψ + mψψ ψγ ψψγµψ . (2.22) T µ − 2 µ Z h i ST can be mapped into the action (2.21) through bosonization [54, 55]. More precisely it can be proved that the duality transformation maps the soliton of the Sine-Gordon 54 Chapter 2. =4 super Yang–Mills theory N
model into the elementary fermion of the Thirring model and the meson states of (2.21) into fermion–anti-fermion bound states. The relation between the coupling constants is
β2 1 = g , 4π 1+ π
so that in this case as well duality establishes a strong/weak coupling correspondence. The crucial rˆole that duality plays in the context of supersymmetric gauge theo- ries relies on this fundamental feature. The kind of duality that is relevant in four dimensional quantum field theories is a generalization of the electric-magnetic duality introduced for the first time by Dirac in [56]. The general features of electric-magnetic duality are reviewed in appendix B, where the case of the Georgi–Glashow (or Yang– Mills–Higgs) model is described. The application to the =4 super Yang–Mills N theory is discussed in this section. There are various reviews on electric-magnetic duality, see in particular [58, 59, 60].
2.4.1 Montonen–Olive and SL(2,Z) duality
Most of the analysis of this section will be carried out for a SU(2) gauge group, but the results can be generalized to larger groups. The classical spectrum of non Abelian gauge theories can contain non-perturbative states associated with solitonic solutions of the equations of motion. The corresponding field configurations are characterized by a topological charge that is interpreted as a magnetic charge, so that these states describe monopoles and in general dyons (states with both electric and magnetic charge) [61]. As discussed in appendix B the Georgi–Glashow model with Lagrangian
1 1 λ 2 = F a F aµν + DµΦaD Φ ΦaΦ v2 , L −4 µν 2 µ a − 4 a − possesses, in the (BPS) limit of vanishing potential, a ‘classical’ spectrum, contain-
ing one scalar massive Higgs field, one massless ‘photon’, massive W ± bosons and
monopoles M ±, that is summarized in the following table 2.4 S-duality 55
Mass (Qe, Qm) Spin Higgs 0 (0,0) 0 Photon 0 (0,0) 1
W ± ve ( e,0) 1 ± M ± vg (0, g) 0 ±
where Qe and Qm are the electric and magnetic charge operators defined as 1 Q = d3x D ΦaBi m v i a Z 1 Q = d3x D ΦaEi , (2.23) e v i a Z i i with Ea and Ba the non Abelian electric and magnetic fields respectively. All of the states saturate the Bogomol’nyi bound [62]
m v Q2 + Q2 , (2.24) ≥ e m p where Q = 4πnm and Q = n e, with n , n Z. Of course quantum corrections m e e e e m ∈ might in principle modify this situation, however the above picture is expected to be a good approximation in the weak coupling limit, e 1. In this limit ≪ 4π m = ev v , m = gv = v v . W ≪ M e ≫ Montonen and Olive have suggested in [37] that a duality transformation exchang- ing “electric” states associated to W ± bosons and “magnetic” states associated to monopoles M ± and at the same time electric and magnetic charges could be an exact symmetry of the full quantum theory in the BPS limit. The exchange of electric and magnetic charges implies, because of the Dirac quantization condition (see appendix B),
4π e g = , −→ e so that the proposed symmetry is intrinsically non-perturbative. The conjecture of [37] is based on the classical spectrum and on the following arguments. First of all the mass formula (2.24) is invariant under duality as can be easily checked. Moreover a semiclassical analysis [63] based on the so called moduli 56 Chapter 2. =4 super Yang–Mills theory N
space approximation (to be discussed in the final subsection) shows that the inter- action between two monopoles vanishes, whereas there is an attractive interaction between monopoles and anti-monopoles. The duality symmetry would require that
the same should be true for the W ± bosons. It can be proved at the semi-classical level that this is indeed the case because the repulsive interaction between equal charge W bosons is exactly cancelled by an identical contribution due to the exchange of scalars, that are massless in the BPS limit. There are some basic problems with the Montonen–Olive proposal, namely
Quantum corrections are expected to spoil the results of the semi-classical anal- • ysis. In particular a non-vanishing potential V (Φ) could be generated.
The elementary W ± bosons have spin one while the monopoles that should be • related to them by the duality have spin zero, making an exact matching of the quantum numbers impossible.
The rˆole of the dyons has been neglected in the previous analysis. • It will be shown that all of these problems can be solved in supersymmetric theo- ries. In particular the =4 supersymmetric Yang–Mills theory is supposed to realize N exactly a generalization of the Montonen–Olive duality called S-duality. This gener- alization emerges when the effects of a non-vanishing vacuum angle are taken into account. In the presence of a θ-term (see the end of section B.2) the Lagrangian of the Georgi–Glashow model can be written
1 1 = Im [τ (F µν + i F µν)(F + i F )] DµΦD Φ , (2.25) L −32π ∗ µν ∗ µν − 2 µ
θ 4π with τ = 2π + i e2 . The electric and magnetic charges of the dyon states saturating eθ 4π the Bogomol’nyi bound are Qe = nee + nm 2π and Qm = nm e respectively. Periodicity in θ requires invariance under
τ τ + b , b Z , (2.26) −→ ∈ which amounts to a shift in the electric charge n n bn , n n , or in matrix e → e − m m → m form n 1 b n e e . (2.27) nm −→ 0 1 nm 2.4 S-duality 57
Furthermore at θ = 0 the duality of Montonen and Olive becomes 1 τ , (2.28) −→ −τ that is equivalent to the exchange of electric and magnetic charges, e g = 4π , → e g e. In Montonen–Olive duality this is supplemented by the exchange of the → − corresponding quantum numbers, Q Q , implying n n , n n . The e ↔ m e → m m → − e latter transformation can be written in matrix form as n 0 1 n e e . (2.29) nm −→ 1 0 nm − It is thus natural to generalize the duality symmetry to the set of transformations gen- erated by (2.26) and (2.28) with arbitrary θ. The resulting group of transformations is SL(2,Z) acting projectively on the parameter τ
aτ + b τ , a,b,c,d Z , ad bc =1 . (2.30) −→ cτ + d ∈ −
The corresponding action on the quantum numbers ne and nm, generated by (2.27) and (2.29), is
n a b n e − e . (2.31) nm −→ c d nm − One can then easily check that the mass formula, written in terms of τ as
2 2 1 1 Re τ ne m 4πv (ne, nm) − 2 (2.32) ≥ Im τ Re τ τ nm − | | is invariant.
2.4.2 S-duality in =4 supersymmetric Yang–Mills theory N The general properties of the =4 supersymmetric Yang–Mills theory previously N discussed allow to give an intuitive explanation of why it is considered the original example of a theory possessing exact S-duality. In the Coulomb phase for each scalar with non-vanishing vev the same construction leading to the ’t Hooft–Polyakov monopole (see the appendix) can be carried on. As a result the classical spectrum contains monopoles and dyons as well. Unlike the case of the Georgi–Glashow model the finiteness of the theory makes the semiclassical analysis of the spectrum sensible even at the quantum level. Moreover the theory has the correct number of fermions in 58 Chapter 2. =4 super Yang–Mills theory N
the adjoint representation to give spin one dyon/monopole configurations through the mechanism sketched in section B.3. More precisely, in the case of a SU(2) gauge group i i being discussed here, creation and annihilation operators a † and a , with i = 1, 2, ± ± associated with the fermionic zero modes can be defined so that the spectrum of zero-energy states, in the charge-one monopole sector, consists of the following states [64]
State Spin (Sz) Ω 0 | i i 1 † a Ω 2 ± | i ± i j a †a+† Ω 0 − | i 1 2 a †a † Ω 1 + + | i 1 2 a †a † Ω -1 − − | i 1 2 i 1 † † † a a a Ω 2 ∓ ∓ ± | i ∓ 1 2 1 2 a+†a+†a †a † Ω 0 − − | i
where Ω is the magnetic charge-one Clifford vacuum. The multiplet generated in | i this way contains sixteen states: eight bosons (six spin 0 scalars and two spin 1 ± vectors) and eight spin 1 fermions. As a result the spin quantum numbers of the ± 2 monopole multiplet exactly match those of the multiplet of elementary fields. Furthermore in extended supersymmetric theories the Bogomol’nyi bound comes as a consequence of the supersymmetry algebra, so that it holds in the full quantum theory. In supersymmetric models a vanishing scalar potential in correspondence of non trivial scalar field configurations is perfectly natural because of the presence of flat directions. In the case of =4 super Yang–Mills this scalar field configuration N corresponds to a generic point of the Coulomb phase. The relation of the BPS limit with the supersymmetry algebra is better under- stood in the =2 case. The =2 supersymmetry algebra possesses two central N N charges which are related to the magnetic and electric charge operators of the theory [65]. Using a Majorana notation, the =2 algebra contains the anticommutator N
Qi , Qj = δijγµ P + δ U ij +(γ ) V ij , i, j =1, 2 , { α β} αβ µ αβ 5 αβ 2.4 S-duality 59
where the central charges have been denoted by U ij and V ij. By expressing U ij and V ij in terms of the fields of the theory Olive and Witten have proved the relations
ij ij ij ij U = ε vQe , V = ε vQm , where Qe and Qm are the electric and magnetic charge operators respectively, that generalize those in equation (2.23). In the previous equation v2 = tr(S2) + tr(P 2) , h i h i where S and P denote the complex Higgs fields of the model. As a consequence of this relation the Bogomol’nyi bound follows directly from the supersymmetry algebra because of the positive definiteness of the operator Q, Q in the rest frame. { } In the =4 case the situation is more involved because there are twelve central N charges. Correspondingly there are six electric and six magnetic charge operators associated to the six scalars, ϕi, of the =4 multiplet N i 1 ~ i Qe = d~σ tr Eϕ v S2 · Z ∞ i 1 ~ i Qm = d~σ tr Bϕ , (2.33) v S2 · Z ∞ where v2 = 6 tr(ϕi)2 . The Bogomol’nyi bound reads i=1h i P m2 v2 (Qi )2 +(Qi )2 ≥ e m and can be saturated only by states annihilated by exactly one half of the super- charges. The BPS states which saturate the limit correspond to short multiplets of the =4 superalgebra and for this reason are protected in the full quantum theory. N The mass of the BPS states is completely determined by their electric and magnetic quantum numbers and is exact because supersymmetry is preserved at the quantum level. All the BPS states in the spectrum can be displayed on six two-dimensional lattices i i (one for each couple of charges Qe and Qm) and lie at points of coordinates eθ 4π = Q + iQ = n e + n + in . Q e m e m 2π m e
Supersymmetry implies that the BPS states in the spectrum can only decay into a couple of BPS states, so that a generic state of mass m is unstable if and only if there exist BPS states of masses m1 and m2 such that
m m + m . ≥ 1 2 60 Chapter 2. =4 super Yang–Mills theory N
Figure 2.1: Lattice of BPS states
It immediately follows that stable BPS states are all and only those which have
integer quantum numbers ne and nm that are coprime. These are the states marked by empty circles in figure 2.1, while the full circles denote unstable states. Thanks to the fact that the BPS states are protected by supersymmetry in the =4 theory a semi-classical analysis of the spectrum is sensible at the quantum level N as well. Some arguments supporting the conjectured S-duality of the =4 super N Yang–Mills theory based on the moduli space approximation are briefly reviewed in the next subsection.
2.4.3 Semiclassical quantization
In appendix B it is discussed how to construct a classical solution of the equations of motion describing a monopole in the Georgi–Glashow model, the aim of this section is to show how the semiclassical quantization of the monopole solution can be used to study the spectrum of the model [63, 66]. The notion of collective coordinates will be introduced starting with the ’t Hooft–Polyakov monopole and then the semiclassical quantization will be applied to the =4 super Yang–Mills case. A readable review N of these topics is provided by [67]. a In the BPS limit the monopole solution satisfies the Bogomol’nyi equation Bi = a DiΦ (i =1, 2, 3, a =1,..., dim G). Given a field configuration solving this equation in general one can find a multi-parameter family of solutions with the same energy. 2.4 S-duality 61
The parameters labeling the degenerate solutions are called moduli or collective co- ordinates of the monopole and the corresponding space moduli space. For the ’t Hooft–Polyakov monopole described in appendix B three collective coor- dinates are associated with the translational invariance of the Bogomol’nyi equation. cl a cl a Given a solution (Ai (~x), Φ (~x)) the field configuration
cl a ~ cl a ~ Ai (~x + X) , Φ (~x + X) , (2.34) where X~ is a constant vector, still satisfies the equation and has the same energy (and magnetic charge). Thus the three coordinates X~ are moduli. A fourth collective coordinates is related to gauge transformations. Gauge invari- ance implies that the functional space of physical field configurations is
= / , C A G where =(Aa, Φa) and is the group of “small” gauge transformations, i.e. gauge A i G transformations that do not reduce to the identity at spatial infinity. “Large”, or “global”, gauge transformations which do not approach the identity at infinity are true symmetries of the theory and so field configurations related by large gauge trans- formations are inequivalent. The BPS solution for the ’t Hooft–Polyakov monopole is symmetric under the diagonal SO(3) subgroup of SO(3) SU(2) , where SO(3) is R× G R the group of spatial rotations and SU(2)G the group of large gauge transformations, but it is not invariant under SU(2)G. As a result by acting with the U(1) subgroup of SU(2)G that is not broken in the Higgs vacuum one generates a new solution with the same energy, so that the fourth collective coordinate is associated with the U(1) group of global ‘electromagnetic’ transformations. By letting X~ depend on time one cl a ~ cl a ~ describes a moving monopole, (Ai (~x+X(t)), Φ (~x+X(t))). Since the Hamiltonian for the system in the temporal gauge A0 = 0 is 1 1 H = T + V = d3x , A˙ aA˙ i + Φ˙ aΦ˙ + d3x , BaBi + D ΦaDiΦ 2 i a a 2 i a i a Z Z h i the time dependence of X~ generates a non vanishing electric field so that the kinetic energy increases while the potential energy remains constant. This is a general feature of a motion in the moduli space. One can then consider a general deformation of a
BPS monopole in the A0 = 0 gauge
a a δAi (~x, t) , δΦ (~x, t) . (2.35) 62 Chapter 2. =4 super Yang–Mills theory N
To keep the potential energy fixed (2.35) must satisfy the linearized Bogomol’nyi equation
ε DjδAk D δΦ+ e[δA , Φcl]=0 (2.36) ijk − i i and the Gauss law constraint
i cl DiδA + e[Φ , δΦ]=0 . (2.37)
The general solution can be written in terms of a single function of time χ(t)
cl δAi = Di χ(t)Φ δΦ = 0 (2.38) δA = D χ(t)Φcl χ˙(t)Φcl . 0 0 − a a ~ If χ is taken to be constant and (δAi , δΦ ) does not contain a translation δ~x = ~x+X, the deformation reduces to a large gauge transformation, so that χ is identified with the fourth collective coordinate. Forχ ˙ = 0 a non-vanishing kinetic energy is gen- 6 erated. Moreover χ is periodic since it parameterizes a compact U(1) (it is a sub-
group of SU(2)G), therefore the moduli space with coordinates (X,~ χ) is topologically = R3 S1. M1 × The previous construction of the moduli space can be generalized to a charge k monopole solution. A rigorous derivation of multimonopole solutions is quite involved [68], the resulting moduli space is 4k-dimensional and roughly speaking the Mk 4k collective coordinates correspond to the locations of the k monopoles and their dyon degrees of freedom. On physical grounds the existence of multimonopole BPS configurations is related to the fact that the repulsion between monopoles of the same charge can be compensated by an attractive interaction mediated by the massless Higgs field, thus allowing the construction of static configurations. a a For a charge k monopole deformations (δαAi , δαΦ ), with α =1,...,k, satisfying the linearized Bogomol’nyi equations and the Gauss law constraint define tangent vectors to . Given the tangent vectors the standard metric on can be con- Mk Mk structed as
= d3x tr δ A δ Ai + δ Φδ Φ . (2.39) Gαβ − α i β α β Z This metric is actually induced by the form of the action for the gauge theory in C cl cl as can be seen by writing the monopole solution as Ai (~x, zα), Φ (~x, zα), where zα’s 2.4 S-duality 63
denote the moduli. The tangent vectors (δαAi, δαΦ) are calculated by differentiating with respect to the zα. In general one has to include a gauge transformation to ensure that (δαAi, δαΦ) are tangent to the moduli space of physical fields, i.e. that the fields cl α cl α A′ = A + δ A δz , Φ′ = Φ + δ Φδz belong to the space . The general form of i i α i α C the tangent vectors is ∂A δ A = i D ǫ α i ∂zα − i α ∂Φ δ Φ = e[Φcl, ǫ ] , α ∂zα − α
β where ǫα(~x, z ) is a suitable gauge parameter that is fixed by requiring that the lin- earized Bogomol’nyi equation (2.36) and the Gauss law constraint (2.37) be satisfied. The action of the theory on the BPS configuration is then obtained after substituting
A (~x, t) Acl(~x, zα(t)) i −→ i A z˙α(t)ǫ (2.40) 0 −→ α Φ(~x, t) Φcl(~x, zα(t)) −→ and reads 1 4πv S = d3xdt tr F clF 0i = dt z˙αz˙β k . (2.41) −2 0i cl Gαβ − e Z Z The action (2.41) is the starting point for the semiclassical quantization which is achieved by studying the corresponding quantum mechanics. In the k = 1 case the moduli are zα =(X,~ χ) and the action becomes
1 4πv ˙ 2 4π 4πv S = dt X~ + χ˙ 2 . (2.42) 2 e ve2 − e Z The resulting wave functions are plane waves
iP~ X~ ineχ ψ = e · e , with n Z, and one can compute the electric charge, Q = ie∂ , so that the e ∈ e − χ quantum mechanical system possesses an infinite tower of states with Qe = nee. For the mass of these states one gets n2ve3 4πv m = e + v[Q2 + Q2 ] , 8π e ≈ e m showing that the Bogomol’nyi bound is saturated. 64 Chapter 2. =4 super Yang–Mills theory N
The preceding arguments can be extended to the case of the =4 super Yang– N Mills theory to find stronger evidence for the exact S-duality of the model. In order to be able to use the previous considerations the simplest case will be considered, namely the gauge group will be taken to be G=SU(2), broken by the vacuum configuration
ϕ2 = ϕ3 = . . . = ϕ6 =0 h i h i h i ϕ1 = Φ , trΦ2 = v2 . h i h i The perturbative states in the bosonic sector of the theory, labeled by quantum
numbers (nm, ne) consist of a massless photon multiplet with charge (0, 0) and massive W boson multiplets with charge (0, 1). The tower of stable BPS states predicted ± ± by S-duality have charges (l,k) (with l and k coprimes) and can be studied within the framework of semiclassical quantization since they can be put in correspondence with certain geometric structures on . Then the non-renormalization properties Mk of the theory make the semiclassical result reliable in the full quantum theory. In the Georgi–Glashow model is parameterized by 4k collective coordinates Mk which are bosonic zero modes. In the =4 supersymmetric Yang–Mills theory there N are fermionic zero modes as well in the monopole background and the index theorem of [71] shows that the number of such zero modes in the charge k sector is precisely 4k. As a consequence the moduli space has fermionic coordinates too and the low energy ansatz to be substituted into the Bogomol’nyi equation is given by (2.40) supplemented by
λ(~x, t) ψ(t)λcl(~x, z (t)) (2.43) ∼ α for the Weyl spinors in the monopole background. In equation (2.43) the ansatz for the spinors λ(~x, t) has been schematically written as the product of a c-number cl λ (~x, zα) satisfying a zero mode equation in the monopole background and a time dependent fermionic collective coordinate ψ(t). The zero modes can be shown to constitute a short multiplet preserving half of the supersymmetries just like is ex- pected for BPS states. The exact ansatz for the fields is rather complicated and was discussed in detail in [69]. A technically involved computation allows to derive the quantum mechanical action to be employed in the semiclassical analysis. The action reads
1 α 1 α β 4πv S = dt z˙αz˙β + iψ γ0D ψβ + R ψ ψγ ψ ψδ k , (2.44) 2 Gαβ t 6 αβγδ − e2 Z 2.4 S-duality 65
α ˙ α α β γ where a Majorana notation is used. In (2.44) Dtψ = ψ +Γβγz˙ ψ is the covariant derivative, with Γα the Christoffel symbols of the metric , and R is the βγ Gαβ αβγδ corresponding Riemann tensor [69]. It can be shown that thanks to the fact that the monopole moduli space is a hyper-K¨ahler manifold, S in equation (2.44) is invariant under the action of eight real supercharges. The quantization of the supersymmetric quantum mechanics (2.44) was performed in [70], where it was shown that the states can be put in one to one correspondence with differential forms on and the Mk Hamiltonian of the system H can be expressed in terms of the exterior derivative acting on forms. This construction leads to a basis of sixteen independent forms that are eigenstates of H in the nm = k = 1 sector. These states constitute a BPS multiplet as can be verified by a computation of the corresponding quantum numbers. The complete wave functions contain a plane wave factor from the bosonic sector of
iP~ X~ ineχ eθ the form e · e so that again one finds a tower of states with Qe = ne + 2π . These states all saturate the Bogomol’nyi bound and therefore are exactly those predicted by S-duality.
For nm = k> 1 the problem is much more complicated. The moduli space is
= R3 (S1 ˜ 0)/Z , Mk × × Mk k where ˜ 0 is a 4(k 1)-dimensional hyper-K¨ahler manifold. The states are tensor Mk − products of forms on R3 S1 with forms on ˜ 0, s = ω, n α 2. The forms × Mk | i | ei⊗| i α can be proved to be unique and much in the same way as in the k = 1 case there | i are sixteen forms ω, ne for all k that give rise to a BPS multiplet. The resulting | i eθ states carry electric charge Qe = nee + nm 2π and saturate the Bogomol’nyi bound. The unique form α for the case k = 2 has been explicitly determined by Sen in [72]. | i Arguments suggesting a possible generalization to k > 2 were given in [73].
2 There are actually further subtleties related to the Zk identification. Chapter 3
Perturbative analysis of =4 supersymmetric Yang–MillsN theory
As discussed in the previous chapter, the =4 supersymmetric Yang–Mills theory N has been proved to be finite at the perturbative level up to three loops and strong arguments have been proposed in order to extend this result to all orders and even non perturbatively. For this reason the theory has long been considered to be rather trivial. However recent developments, within the context of the correspondence with type IIB superstring theory on anti de Sitter space, that will be studied in the final chapter, have shown that it is actually an extremely interesting superconformal field theory, possessing very peculiar properties. Even in perturbation theory a careful analysis allows to point out various problems mostly related to the choice of gauge. Both in the component field formulation and using =1 superfields the gauge fixing procedure appears very subtle. As will be N discussed in detail in both cases one finds divergences in off-shell Green functions. In the component formulation the propagators of the elementary fields are ul- traviolet divergent in the Wess–Zumino (WZ) gauge and these infinities are exactly cancelled when the contributions of the gauge-dependent fields, that are put to zero in the WZ gauge, are taken into account. The choice of the WZ gauge, that is almost unavoidable in explicit computations, introduces divergences that require a wave function renormalization. This is a general result that applies to theories with less supersymmetry as well, but appears particularly dramatic in the =4 case be- N 66 67
cause the theory is finite, so that the divergences cannot be reabsorbed through a redefinition of the bare parameters. Different problems emerge in the formulation of the theory in terms of =1 super- N fields. Almost all the calculations showing the vanishing of the quantum corrections to two- and three-points functions that are presented in the literature were performed in the supersymmetric generalization of the Fermi–Feynman gauge (α=1 in the no- tation of section 1.6). With a different choice of gauge two and three point functions develop infrared singularities of difficult interpretation, leading to the result that the choice α=1 is somehow privileged. This conclusion was proposed for the first time in [74] and then it was discussed in the case of the =4 theory in [75]; however no N possible explanation was proposed. Unlike those of the elementary fields the correla- tion functions of gauge invariant composite operators, that play a crucial rˆole in the correspondence with AdS type IIB supergravity/superstring theory, should not suffer from problems related to the gauge fixing. Also, it will be discussed, in the superfield formulation, the effect of the intro- duction of a mass term for the chiral superfields, which breaks supersymmetry from =4 down to =1. It will be shown that this deformation of the model does not N N modify the ultraviolet properties of the original theory. This result was first proposed in [76], where it was proved that the inclusion of mass terms for the (anti) chiral su- perfields does not generate divergent corrections to the effective action. It will be argued that this statement can be reinforced, showing that, at least at one loop, no new divergences, not even corresponding to wave function renormalizations, appear as a consequence of the addition of the mass terms. This result supports the claim put forward in [77, 78], where the ‘mass deformed’ =4 theory was proposed as a N supersymmetry-preserving regularization scheme for a class of =1 theories. It will N also be argued that the same approach should work in the case of =2 theories. N All of this problems are examined in detail in this chapter. Section 3.1 deals with the difficulties introduced by the choice of the Wess–Zumino gauge when the com- ponent formalism is used. The subsequent sections report calculations performed in the =1 superfield formalism of two-, three- and four-point functions. The material N presented in this chapter is an original and systematic re-analysis of the perturbative properties of =4 supersymmetric Yang–Mills theory, the treatment follows closely N [79]. 68 Chapter 3. Perturbative analysis of =4 SYM theory N
3.1 Perturbation theory in components: problems with the Wess–Zumino gauge
Throughout this chapter calculations will be carried on in Euclidean space-time and unless otherwise stated the gauge group will be taken to be SU(N). To discuss per- turbation theory in components in this section the formulation of equation (2.11), with a SU(3) U(1) subgroup of the SU(4) R-symmetry group manifest, will be em- × ployed. To correctly deal with the lower components of the vector superfield V , that are put to zero in the Wess–Zumino gauge, it is however useful not to eliminate the auxiliary fields F and D through their equations of motion, but rather keep them in the action: in the computation of Green functions the corresponding x-space prop- agators are simply δ-functions. The complete expression for the vector superfield V
was given in equation (1.24) and contains beyond the physical fields, Aµ and λ, and the auxiliary field, D, the ‘gauge-dependent’ fields C, χ and S. Note that the scalar field C and the spinor χ have “wrong” physical dimension, resulting in non standard
free propagators, as will be shown. The non Abelian field strength superfield Wα is
∞ 1 V V (k) W = DDe− D e = W , (3.1) α −4 α α Xk=1 where 1 W (1) = DDD V α −4 α (3.2) 1 W (2) = DD[V,D V ] α 8 α
(k) and the terms Wα , with k 3, contain k factors of V and vanish in the WZ gauge. ≥ The Euclidean action in =1 superfields can be written N
(E) 1 4 4 I 1 1 (1)α (1) 1 (1) (1)α ˙ S = tr d xd θ ΦI† Φ W Wα δ(θ)+ W α˙ W δ(θ)+ dr − − 4g2 4 4 Z 1 2 2 I 1 (1)α (2) 1 (2)α (2) D VD V g[Φ† V ]Φ W W + W W δ(θ)+ −8α − I − 8g2 α 16g2 α 1 (1) (2)α ˙ 1 (2) (2)α ˙ 1 2 I + W W + W W δ(θ) g [V, [V, Φ† ]]Φ + . . . , (3.3) 8g2 α˙ 16g2 α˙ − 2 I where the gauge fixing term has been included, whereas no ghost term is displayed since it will not be relevant for the computations to be discussed in this section. In equation (3.3) the dots denote terms of higher order in V , which do not contribute 3.1 Perturbation theory in components 69
to the Green functions that will be considered and that will be suppressed from now on. In the following calculations the Fermi–Feynman gauge, α=1, will be used. With this choice one obtains 1
1 4 4 2 I 1 (1)α (2) 1 (2)α (2) S = tr d xd θ V V ΦI† Φ W Wα + W Wα δ(θ)+ dr − − 8g2 16g2 Z h i 1 (1) (2)α ˙ 1 (2) (2)α ˙ I 1 2 I + W W + W W δ(θ) g[Φ† V ]Φ + g [V, [V, Φ† ]]Φ , 8g2 α˙ 16g2 α˙ − − I 2 I where from the definition (3.2) it follows
1 i W (1) = iλ + δβD δβ2C (σµσν) β (∂ A ∂ A ) θ + α − α α − 2 α − 2 α µ ν − ν µ β µ α˙ +θθσαα˙ ∂µλ . (3.4)
(2) The explicit form of Wα will not be necessary for the moment (see however equation (3.18)). Expansion of the action in components using the complete expression of V gives
S = S0 + Sint .
I The free action S comes from the terms V 2V and Φ† Φ and reads 0 − I
4 a µ I a µ I a I 2 a S0 = d x (∂µϕI†)(∂ ϕa)+ ψI σ (∂µψa)+ FI †Fa + Sa† S + Z 1 1nh 1 1 1 i 1 Ca2D Da2C Ca22C + χa2λ + χa2λ + λa2χ + −2 a − 2 a − 2 a 2 a 2 a 2 a 1 a 1 1 1 + λ 2χ + χa2σµ(∂ χ )+ χa2σµ(∂ χ )+ (∂ Aa)(∂ A ) . (3.5) 2 a 2 µ a 2 µ a 2 µ ν ν aµ
The interaction part Sint contains an infinite number of terms. The propagators of the fermion ψ and of the scalar ϕ belonging to the =1 chiral multiplet will now be N computed at one loop. The terms that are relevant for these calculations come from 2 the expansion of Φ†V Φ and Φ†V Φ in the superfield action. The latter generates tadpole type diagrams in the propagator ϕ†ϕ , but not in ψψ . The interaction h i h i 1From now on the superscript E will be suppressed. Euclidean signature is understood unless otherwise stated. 70 Chapter 3. Perturbative analysis of =4 SYM theory N
part of the action to be considered is
4 1 a b cI 1 a b cI 1 a b µ cI S = d x igf φ †D φ + φ †(2C )φ (∂ φ †)C (∂ φ )+ int abc 2 I 2 I − 2 µ I Z i a b µ cI µ a b cI i a b cI a b cI + φ †A (∂ φ ) (∂ φ †)A φ + φ †S †F F †S φ + 2 I µ − I µ √2 I − I a b cI i a b cI a b cI i a b cI a b cI F †C F + ψ λ φ φ †λ ψ + F †χ ψ ψ χ F + − I √2 I − I √2 I − I i a b µ cI a µ b cI 1 b a µ cI b a µ cI φ †(∂µχ )σ ψ + ψ σ (∂µχ )φ + C ψ σ (∂µψ ) C (∂µψ )σ ψ + −√2 I 2 I − I i a µ cI b √2 IJK a b c a b c aI bJ cK ψ σ ψ A gf ε φ †φ †F † φ †ψ ψ + ε φ φ F + −2 I µ − 3! abc I J K − I J K IJK 2 g a φaI ψbJ ψcK f f e CbψdI σµψ Ac + − − 2 abe cd − I µ i c d b aI i dI µ a dI µ a b c a b c +(χ ψ )(χ ψ )+ ∂ ψ σ ψ ψ σ ∂ ψ C C + ϕ † C D + I 2 µ I − µ I I 1 c b c i µ c b c i µ c b c + 2C χ λ + σ ∂ χ χ λ + σ ∂ χ + S †S 2 − 2 µ − 2 µ − 1 bc cµ dI i b cµ a dI a dI 1 b c a dI A A ϕ + C A ϕ †(∂ ϕ ) (∂ ϕ †)ϕ + C C ϕ †(2ϕ )+ −2 µ 2 I µ − µ I 4 I a dI i b µ c a dI a dI +(2ϕ †)ϕ + χ σ χ ϕ †(∂ ϕ ) (∂ ϕ †)ϕ . (3.6) I 2 I µ − µ I The quadratic part of the action, S0, can be written in a more compact form intro- ducing the notation χa(x) a a C (x) χ (x) a(x)= , a(x)= , (3.7) B a F a D (x) ! λ (x) a λ (x) so that
4 a µ I a µ I a I S0 = d x (∂µϕI†)(∂ ϕa)+ ψI σ (∂µψa)+ FI †Fa + Z nh i a 1 a µ T a T a + S†2S A A + M + N , (3.8) a − 2 µ a Ba B Fa F where 0 2σµ∂ 2 0 µ − 1 22 2 1 2σµ∂ 0 0 2 M = − − , N = µ − . (3.9) 2 2 0 2 2 0 0 0 − − 0 2 0 0 − 3.1 Perturbation theory in components 71
Inverting the kinetic matrices M and N one gets the free propagators. From (3.9) it follows 1 0 0 0 −2 1 1 0 0 0 0 −2 1 2 1 M − = 1 , N − = 1 σµ∂ , (3.10) µ 2 1 2 0 0 2 − − µ 1 σ ∂µ 0 2 2 0 − so that the free propagators are J, b I, a b I b I δaδJ bI ϕ †(x)ϕ (y) = δ(x y) = ∆ (x y) −→ h J a ifree − 2 − aJ −
b a b b δa b S †(x)S (y) = δ(x y)=2∆ (x y) −→ h a ifree 2 − a −
b a δb Cb(x)D (y) = a δ(x y) = ∆b (x y) −→ h a ifree 2 − a −
b a Db(x)D (y) = δbδ(x y) −→ h a ifree − a −
J, b I, a b I I b F †(x)F (y) = δ δ δ(x y) −→ h J a ifree J a −
µ, b ν, a δ δb Ab (x)A (y) = µν a δ(x y) = ∆b (x y) −→ h µ aν ifree − 2 − aµν −
α,b β,a εαβδb χbα(x)λβ(y) = a δ(x y)= Rbαβ(x y) −→ h a ifree − 2 − a −
˙ α˙ β˙ b α,˙ b β, a β˙ ε δ bα˙ β˙ χbα˙ (x)λ (y) = a δ(x y)= R (x y) −→ h a ifree − 2 − a −
b µ α,˙ b α, a bα˙ δaσ ∂µ λ (x)λα(y) = αα˙ δ(x y)= −→ h a ifree 2 − b =S (x y) aαα˙ − 72 Chapter 3. Perturbative analysis of =4 SYM theory N
b J µ α,I,b˙ α, J, a bα˙ δaδI σ ∂µ ψ (x)ψαJ (y) = αα˙ δ(x y)= −→ h I a ifree 2 − bJ =S (x y) aIαα˙ − In conclusion beyond the ordinary propagators for the physical fields, ϕ, ψ, λ and
Aµ, and those for the auxiliary fields, F and D, one further obtains the propagators
S†S , CD , χλ and χλ . The latter are not present in the Wess–Zumino gauge. h i h i h i h i
3.1.1 One loop corrections to the propagator of the fermions belonging to the chiral multiplet
The one-loop correction to the propagator of the fermions ψI in the chiral multiplet is the simplest to compute. In the WZ gauge there are three contributions at the one loop level, that will be shown to lead to a logarithmically divergent result. From the action (3.5), (3.6), with C=χ=S=0 one obtains the following three diagrams
AµAν
a ψ (x) ψI (y) AaI (x; y) J b −→ bJ ψψ ψψ
a ψ (x) ψI (y) BaI (x; y) J b −→ bJ
ϕϕ†
λλ
a ψ (x) ψI (y) CaI (x; y) J b −→ bJ
ϕ†ϕ
The three contributions depicted above can be easily evaluated and give the results
1 αβaI˙ AααaI˙ (x; y) = g2f d f l d4x d4x ∆me(x x ) S (x x )σν bJ −2 ef mn 1 2 νµ 2 − 1 lL − 2 βγ˙ · Z γγnL˙ n βαfK˙ h S (x x )σµ S (x y) , (3.11) · dK 2 − 1 γβ˙ bJ 1 − io 3.1 Perturbation theory in components 73
1 BααaI˙ (x; y) = g2εLMN εPQ f f f d4x d4x ∆ld (x x ) bJ −2 R de lmn 1 2 PL 2 − 1 · Z αβaI˙ βαnR˙ n S (x x )Sem (x x )S (x y) , (3.12) · fN − 2 ββMQ˙ 2 − 1 bJ 1 − h io CααaI˙ (x; y) = g2f f f n d4x d4x ∆lK (x x ) bJ − de lm 1 2 fL 2 − 1 · Z αβaI˙ n βαdL˙ S (x x )Sem(x x )S (x y) . (3.13) · nK − 2 ββ˙ 2 − 1 bJ 1 − Taking the Fourier transformh one obtains io
αβ˙ βα˙ 4 ˜ααaI˙ i 2 2 a I ˜ λ ˜ d k 1 AbJ (p)= g N(N 1)δb δJ S (p)σββ˙ pλS (p) 4 2 2 , 2 − " (2π) k + p k p # Z 2 − 2 B˜ααaI˙ (p)= C˜ααaI˙ (p)= A˜ααaI˙ (p) , bJ − bJ bJ where αα˙ σαα˙ pµ S˜ (p)= i µ . − p2 In conclusion the one-loop correction to the fermion propagator in the Wess–Zumino gauge turns out to be logarithmically divergent
ααaI˙ 1 loop,WZ ααaI˙ ααaI˙ ααaI˙ ψψ − = A˜ (p)+ B˜ (p)+ C˜ (p)= h bJ iFT bJ bJ bJ αβ˙ βα˙ 4 i 2 2 a I ˜ λ ˜ d k 1 = g N(N 1)δb δJ S (p)σββ˙ pλS (p) 4 2 2 . (3.14) 2 − " (2π) k + p k p # Z 2 − 2 Equation (3.14) shows that a logarithmically divergent wav e function renormalization is required in the WZ gauge. This divergence will now be proved to be a gauge artifact due to the choice of the WZ gauge. In fact if the fields C, χ and S are included two further contributions must be taken into account, which correspond to the diagrams
λχ
a ψ (x) ψI (y) DaI (x; y) J b −→ bJ
ϕ†ϕ λχ
a ψ (x) ψI (y) EaI (x; y) J b −→ bJ
ϕ†ϕ 74 Chapter 3. Perturbative analysis of =4 SYM theory N
The contribution of these two diagrams exactly cancels the divergence in equa- tion (3.14) giving a net vanishing one-loop result for the propagator. In fact the computation of D and E gives
1 DααaI˙ (x; y)= g2f f f n d4x d4x ∆lK (x x ) bJ −2 de lm 1 2 fL 2 − 1 · Z αγaI˙ βαdL˙ S (x x )σµ ∂(2)Rγne˙ (x x ) S (x y) , (3.15) · mK − 2 γγ˙ µ β˙ 2 − 1 bJ 1 − h io ααaI˙ ααaI˙ EbJ (x; y)= DbJ (x; y) .
In momentum space one has
αβ˙ βα˙ 4 ˜ ααaI˙ i 2 2 a I ˜ λ ˜ d k 1 DbJ (p)= g N(N 1)δb δJ S (p)σββ˙ pλS (p) 4 2 2 , −4 − " (2π) k + p k p # Z 2 − 2 ˜ααaI˙ ˜ ααaI˙ EbJ (p)= DbJ (p) .
In conclusion summing up all the terms gives
ααaI˙ 1 loop ψψ − = h bJ iFT ˜ααaI˙ ˜ααaI˙ ˜ααaI˙ ˜ ααaI˙ ˜ααaI˙ = AbJ (p)+ BbJ (p)+ CbJ (p)+ DbJ (p)+ EbJ (p)=0 , (3.16)
so that the one-loop correction to the ψψ propagator, if the Wess–Zumino gauge is h i not exploited to perform the calculation, is zero, as expected in =4 super Yang– N Mills theory. Notice that the insertion of tadpoles such as
in a diagram gives a vanishing result because all the propagators are diagonal in colour space, so that the tadpole contains a factor δ f abc 0. The same is true also ab ≡ for diagrams in =1 superspace that will be discussed in subsequent sections. N 3.1.2 One loop corrections to the propagator of the scalars belonging to the chiral multiplet
The calculation of the propagator ϕ†ϕ is more complicated because more diagrams h i 2 are involved. In particular tadpole type graphs, coming from the term Φ†V Φ in the 3.1 Perturbation theory in components 75
action, are present. However the result is analogous to what was found for the ψψ h i propagator: in the WZ gauge the one-loop correction is logarithmically divergent, requiring a wave function renormalization, but when the contributions neglected in the WZ gauge are considered the total one-loop result is vanishing.
The diagrams contributing to ϕ†ϕ at the one-loop level in the Wess–Zumino h i gauge are the following
AµAν
a I aI ϕ †(x) ϕ (y) A (x; y) J b −→ bJ
ϕ†ϕ
λλ
a I aI ϕ †(x) ϕ (y) B (x; y) J b −→ bJ ψψ
ψψ
a I aI ϕ †(x) ϕ (y) C (x; y) J b −→ bJ ψψ 76 Chapter 3. Perturbative analysis of =4 SYM theory N
AµAν
a I aI ϕ †(x) ϕ (y) D (x; y) J b −→ bJ
ϕ†ϕ
a I aI ϕ †(x) ϕ (y) E (x; y) J b −→ bJ DD
F F †
a I aI ϕ †(x) ϕ (y) F (x; y) J b −→ bJ
ϕϕ†
The last three diagrams are all tadpole corrections since the free propagators for the auxiliary fields F and D are δ-functions, so that the corresponding lines shrink to a point. Each of these diagrams is quadratically divergent. There is also a quadratically divergent contribution coming from the two diagrams B(x, y) and C(x, y) and the sum of all these terms exactly vanishes. The sum of the previous diagrams gives a final result that is logarithmically divergent and in momentum space is schematically of the form
4 aI 1 loop,WZ 2 I a 2 d k 1 (ϕ†ϕ)bJ FT− g δJ δb p . (3.17) h i ∼ (2π)4 k + p 2 k p 2 Z 2 − 2
Just like in the case of ψψ this divergence corresponds to a wave function renor- h i malization for the field ϕ(x).
Taking into account the contribution of the fields C, χ and S there are eight more diagrams to be calculated 3.1 Perturbation theory in components 77
χλ
a I aI ϕ †(x) ϕ (y) G (x; y) J b −→ bJ ψψ
λχ
a I aI ϕ †(x) ϕ (y) H (x; y) J b −→ bJ ψψ
CD
a I aI ϕ †(x) ϕ (y) L (x; y) J b −→ bJ ψψ
F †F
a I aI ϕ †(x) ϕ (y) M (x; y) J b −→ bJ
SS†
λχ
a I aI ϕ †(x) ϕ (y) N (x; y) J b −→ bJ λχ
a I aI ϕ †(x) ϕ (y) P (x; y) J b −→ bJ
CD
a I aI ϕ †(x) ϕ (y) Q (x; y) J b −→ bJ 78 Chapter 3. Perturbative analysis of =4 SYM theory N
SS†
a I aI ϕ †(x) ϕ (y) R (x; y) J b −→ bJ
Single diagrams contain quadratically divergent terms that cancel in the sum leav- ing a total contribution that is again logarithmically divergent. This non-vanishing correction is exactly what is needed to cancel the divergence in (3.17). As a result
the complete one-loop correction to the ϕ†ϕ propagator is zero, whereas in the h i Wess–Zumino gauge a wave function renormalization is required.
Notice that the situation induced by the choice of the WZ gauge is a feature common to every supersymmetric gauge theory, since the contribution of the gauge- dependent fields C, χ and S is in general logarithmically divergent. However in theo- ries with less supersymmetry a logarithmic wave function renormalization is usually necessary in any case, so that this effect is completely irrelevant. On the contrary it becomes important in the =4 Yang–Mills theory and in general in finite theo- N ries. Of course the divergences encountered here are gauge artifacts and disappear in gauge invariant correlation functions. Examples of computations of correlators of gauge invariant operators, in which the choice of the WZ gauge does not lead to this kind of problems, will be considered within the discussion of the correspondence with AdS type IIB superstring theory in the final chapter. Note also that the divergences encountered in the propagators are zero on-shell.
The calculations described here seem to suggest the possibility of constructing improved Feynman rules in which the effect of the gauge dependent fields, neglected in the WZ gauge, is dealt with by a suitable redefinition of the free propagators. However this program proves extremely complicated when the propagators of fields in the vector multiplet or three- and more-point functions are considered. The calculation of these correlation functions requires, already at the one-loop level, many other (1) (2) terms to be included in the action Sint. In particular terms coming from W W and W (2)W (2) must be considered. The calculation of W (2) without the simplifications 3.2 Propagators in the =1 formulation 79 N introduced by the Wess–Zumino gauge is rather lengthy and gives
a(2) i a b c 1 µ αb˙ c i µ αb˙ c 1 b c W = f iC λ σ χ ∂ C σ χ A + S †χ + α −2 bc − α − 2 αα˙ µ − 2 αα˙ µ √ α 2 b c β β b c i µ ν β b c c + S †S δ + δ C D (σ σ ) C (∂ A ∂ A )+ α α − 2 α µ ν − ν µ i 1 c (σµσν) β(Ab ∂ Cc + Ab ∂ Cc) δ βCb2Cc δ βχbλ + −2 α ν µ µ ν − 2 α − α 1 + (σµσν ) β∂ Cb∂ Cc χbβλc λbβχc iσµ χbα˙ ∂ χcβ+ 2 α µ ν − α − α − αα˙ µ 1 cα˙ +iεγβσµ ∂ χbχc + (σµσν) βAb Ac θ + Cbσµ ∂ λ + γβ˙ µ α 2 α ν µ β αα˙ µ 1 1h iχb Dc (σµσν) βχb (∂ Ac ∂ Ac ) Ab (σµσν) β∂ χc + − α − 4 α β µ ν − ν µ − 2 ν α µ β 1 bµ c i b µ ν β c i µ αb˙ c b c + ∂µA χ ∂ν C (σ σ )α ∂µχ σ ∂µχ S + √2S λ + 2 α − 2 β − √2 αα˙ α cα˙ iAb σµ λ θθ (3.18) − µ αα˙ i
Substituting (3.18) into Sint results in a number of relevant interaction terms of the order of 100, making this formulation totally impractical in explicit calculations. In conclusion perturbation theory in components requires almost unavoidably the WZ gauge, but the latter introduces divergences in gauge dependent quantities. In the following sections the superfield formalism, that allows to avoid these problems, will be employed. However as will be shown new difficulties related to the gauge fixing emerge.
3.2 Perturbation theory in =1 superspace: prop- N agators
The difficulties encountered in the previous section in the calculation of gauge depen- dent quantities, because of the divergences introduced by the Wess–Zumino gauge, can be overcome using the =1 superfield formulation of the model. N In chapter 2 it has already been pointed out how the =1 superfield formalism N has proved to be a powerful tool in the proof of the finiteness of the theory up to three loops. In this approach there is no particular difficulty in working without fixing the WZ gauge. If one does not choose to work in the WZ gauge the action is non polynomial, however a finite number of terms is relevant at each order in perturbation 80 Chapter 3. Perturbative analysis of =4 SYM theory N
theory, so that only at very high order the Wess–Zumino gauge introduces significant simplifications. The aim of this section is to show that there are other subtleties related to the gauge fixing, even if one does not work in the WZ gauge. The formulation of the theory in =1 superspace that will be employed is a ‘mass N deformed’ version of that given by equation (2.10). In other words the action that will be considered is that of equation (2.10) plus mass terms for the (anti) chiral superfields
1 S = d4xd4θ mδ ΦI (x, θ, θ)ΦJa(x, θ, θ)δ(θ)+ m − 2 IJ a Z 1 IJ a + m∗δ Φ† (x, θ, θ)Φ† (x, θ, θ)δ(θ) . (3.19) 2 I Ja The addition of this term to the action breaks supersymmetry down to =1. In [76] it N was argued, by means of dimensional arguments, that this term should not affect the ultraviolet properties of the =4 theory. More precisely the statement of [76] is that N no divergences appear in gauge invariant quantities, so that no divergent contribution to the quantum effective action is generated perturbatively. As a result the model obtained in this way would be an example of a finite =1 theory. The inverse N construction, in which one deforms a =1 model to =4 super Yang–Mills plus a N N mass term, has been proposed in [77, 78] as a regularization procedure preserving supersymmetry. The calculations presented in this section will show that in the presence of the term (3.19) no divergence is generated, at the one-loop level, in the two-, three- and four-point Green functions that are computed. The results presented suggest the possibility of reinforcing the conclusions of [76]; namely the =4 theory N augmented with (3.19) appears to be finite in the sense that no divergences appear in the total n-point irreducible Green functions, at least at one loop. In particular no wave function renormalization is required. Notice that this result is what is actually necessary for the consistency of the approach advocated in [77, 78]. The complete action is 2
S = d4xd4θ V a(x, θ, θ) [ 2P ξ(P + P )2] V (x, θ, θ)+ − T − 1 2 a Z a I 1 I Ja +Φ† (x, θ, θ)Φ (x, θ, θ) mδ Φ (x, θ, θ)Φ (x, θ, θ)δ(θ)+ I a − 2 IJ a 1 IJ a a b mδ Φ† (x, θ, θ)Φ† (x, θ, θ)δ(θ)+ igf Φ† (x, θ, θ)V (x, θ, θ) (3.20) −2 I Ja abc I · 2In the following the mass parameter m will be taken to be real for simplicity of notation. 3.2 Propagators in the =1 formulation 81 N
Ic 1 2 e a b c Id Φ (x, θ, θ) g f f Φ† (x, θ, θ)V (x, θ, θ)V (x, θ, θ)Φ (x, θ, θ)+ · − 2 ab ecd I i 2 gf D DαV a(x, θ, θ) V b(x, θ, θ) D V c(x, θ, θ) + −4 abc α 1 h i 2 g2f ef V a(x, θ, θ) DαV b(x, θ, θ) D V c(x, θ, θ) D V d(x, θ, θ) + −8 ab ecd α √2 h i + . . . gf abc ε ΦI (x, θ, θ)ΦJ (x, θ, θ)ΦK (x, θ, θ)δ(θ)+ − 3! IJK a b c IJK a + ε Φ†Ia(x, θ, θ)Φ†Jb(x, θ, θ)Φ†Kc(x, θ, θ)δ(θ) + C′a(x, θ, θ)C (x, θ, θ)+
a i a a C′a(x, θ, θ)C (x, θ, θ) + gfabc C′ (x, θ, θ)+ C′ (x, θ, θ) − 2√2 · b c c 1 2 e a a V (x, θ, θ) C (x, θ, θ)+ C (x, θ, θ) g f f C′ (x, θ, θ)+ C′ (x, θ, θ) · − 8 ab ecd · d V b(x, θ, θ)V c(x, θ, θ) Cd(x, θ, θ)+ C (x, θ, θ) + . . . , (3.21) · where the dots stand for terms that are not relevant for the considerations of this chapter. Notice that in the action (3.21) a gauge fixing term corresponding to a family 1 of gauges parameterized by α = ξ has been introduced. It will now be shown, by explicitly computing the propagators of both the chiral and the vector superfields, that the supersymmetric generalization of the Fermi–Feynman gauge, corresponding to α = 1, is somehow privileged (see also [75, 80]), since any other choice of the parameter α leads to infrared divergences in Green functions.
3.2.1 Propagator of the chiral superfield
The propagator of the chiral superfield is the simplest Green function to compute. The calculation will be reported in detail in order to illustrate the superfield tech- nique. It follows from the action (2.10) that there are three diagrams contributing to the propagator ΦΦ† at the one loop level. The one-particle irreducible parts of these h i diagrams will be directly evaluated in momentum space using the super Feynman rules of section 1.6. The convention employed is that all momenta are taken to be incoming. The diagrams are the following 82 Chapter 3. Perturbative analysis of =4 SYM theory N
VV
a I Φ †(z) Φ (z′) A(z; z′) J b −→
ΦΦ†
Φ†Φ
a I Φ †(z) Φ (z′) B(z; z′) J b −→
Φ†Φ
VV
a I Φ †(z) Φ (z′) C(z; z′) J b −→
where z =(x, θ, θ) and the notation for the internal propagators is
I, a J, b aI b I a 1 Φ (z)Φ †(z′) = δ δ δ (z z′) −→ h J ifree J b 2 + m2 8 −
a b a a δb V (z)V (z′) = [1+(α 1)(P + P )]δ (z z′) −→ h b ifree − 2 − 1 2 8 −
P and P in the VV propagator are the projectors (1.63). Moreover in the presence 1 2 h i of mass terms for Φ and Φ† there are free propagators ΦΦ and Φ†Φ† , that will h i h i enter the calculation of the vector superfield propagator at one loop
I, a J, b 2 aI J IJ a m D Φ (z)Φ (z′) = δ δ δ (z z′) −→ h b ifree − b 4 2(2 + m2) 8 −
I, a J, b 2 a a m D Φ †(z)Φ† (z′) = δ δ δ (z z′) −→ h I bJ ifree − IJ b 4 2(2 + m2) 8 −
Using these free propagators and the rules of section 1.6, with the vertices read 3.2 Propagators in the =1 formulation 83 N from the action (3.21), the three contributions can be evaluated as follows
4 I cf d k 4 4 a 1 2 δJ δ δ(1, 2) A˜(p) = d θ d θ igf Φ †(p, θ , θ ) D (2π)4 1 2 acd I 1 1 −4 1 (p k)2 + m2 · Z − ←− de 1 2 2 2 2 2 δ δ(1, 2) bJ D1 1+ γ(D1D1 + D1D1) 2 igfefbΦ ( p, θ2, θ2) , · −4 − k − ) where γ =(α 1) and the compact notation δ(1, 2) = δ (θ θ )δ (θ θ ) has been − 2 1 − 2 2 1 − 2 introduced. The computation uses the properties of the δ-function, which imply for example [23, 4]
←− ←− D1αδ(1, 2) = δ(1, 2)D2α , D1˙αδ(1, 2) = δ(1, 2)D2˙α , − ←− ←− − ←− ←− 2 2 2 2 D1˙αD1αδ(1, 2) = δ(1, 2)D2˙αD2α , D1D1δ(1, 2) = δ(1, 2)D2D2 , and integrations by parts on the Grassmannian variables in order to remove the D and D derivatives from one δ, so that one θ integration can be performed immediately. In this way one obtains an expression that is local in θ as is expected from the =1 N non-renormalization theorem. For A˜(p) one obtains
2 4 2 2 a I 1 d k 4 4 1 A˜(p) = g N(N 1)δ δ d θ d θ Φ† (p, 1) − − b J 4 (2π)4 1 2 k2[(p k)2 + m2] aI · Z − n 2 2 2 ΦbJ ( p, 2) D D2δ(1, 2) 1+ γ(D2D + D D2)δ(1, 2) = A˜ (p)+ A˜ (p) . · − 1 1 1 1 1 1 1 2 h i h io The first term is trivially calculated using [23, 4]
4 2 2 d θ D D δ(θ θ′) δ(θ θ′)=16 − − Z h i 4 m n d θ D D δ(θ θ′) δ(θ θ′)=0 if (m, n) = (2, 2) (3.22) − − 6 Z and gives
4 2 2 a I d k 4 1 bJ A˜ (p)= g N(N 1)δ δ d θ Φ† (p,θ, θ)Φ ( p,θ, θ) . 1 − − b J (2π)4 k2[(p k)2 + m2] aI − Z − n (3.23)o
In the computation of the second term one must use the (anti) commutators of covariant derivatives, see appendix A, which in particular imply [23, 4]
2 2 2 2 D D2D = 162D D2D D2 = 162D2 . (3.24) 84 Chapter 3. Perturbative analysis of =4 SYM theory N
Then integration by parts analogously gives
4 2 2 2 2 a I d k 4 [(p k) + p ] bJ A˜ (p)= γg N(N 1)δ δ d θ − Φ† (p,θ, θ)Φ ( p,θ, θ) . 2 − b J (2π)4 k4[(p k)2 + m2] aI − Z − n (3.25)o In conclusion from the first diagram one obtains two contributions, the first propor- tional to γ and the second independent of it. The second diagram gives one single γ-independent contribution. The Feynman rules give
4 d k 4 4 √2 I a B˜(p)= d θ d θ gε f Φ† (p, 1) (2π)4 1 2 − 3! KL acd I · Z ( ! ce K ←− d L 1 δ δ δ(1, 2) 1 2 δ δ δ(1, 2) D 2 M D f N · −4 1 (p k)2 + m2 −4 2 k2 + m2 · − " #
√2 MN f bJ εJ fbe Φ ( p, 2) . · − 3! ! − )
Proceeding exactly as for A˜1 one obtains
4 bJ d k Φ† (p,θ, θ)Φ ( p,θ, θ) B˜(p)= g2N(N 2 1)δaδI d4θ aI − . (3.26) − b J (2π)4 (k2 + m2)[(p k)2 + m2] Z ( − ) From the last diagram one gets
4 2 d k 4 g d a C˜(p)= d θ f f Φ †(p, 1) (2π)4 1 − 2 ac deb I · Z ce 2 2 δ δ(1, 1) 1+ γ(D D2 + D2D ) ΦbJ ( p, 1) . · − 1 1 1 1 k2 − The only non-vanishing contribution comes from the term in which the projection operators act on the δ-function, since δ (θ θ) = 0, so that 4 − 4 2 2 b I d k 4 1 a J C˜(p)= γg N(N 1)δ δ d θ Φ †(p,θ, θ)Φ ( p,θ, θ) , (3.27) − − a J (2π)4 k4 I b − Z n o from which one sees that the C contribution is absent in the gauge α = 1, i.e. γ = 0. Putting the various corrections together gives the following result. The sum of
the terms A˜1(p) and B˜(p), which does not depend on γ, is
4 2 2 b I d k 4 a J A˜ (p)+ B˜(p) = g N(N 1)δ δ d θ Φ †(p,θ, θ)Φ ( p,θ, θ) 1 − a J (2π)4 I b − · Z 1 h 1 i . · k2 [(p k)2 + m2] − (k2 + m2) [(p k)2 + m2] − − 3.2 Propagators in the =1 formulation 85 N
This is finite and exactly vanishes for m = 0, i.e. in the limit in which the =4 N theory is recovered.
The sum of the terms A˜2(p) and C˜(p) is proportional to γ and reads
4 2 2 b I d k 4 a J A˜ (p)+ C˜(p) = γg N(N 1)δ δ d θ Φ †(p,θ, θ)Φ ( p,θ, θ) 2 − a J (2π)4 I b − · Z (p k)2 1 h p2 i − + . · k4 [(p k)2 + m2] − k4 k4 [(p k)2 + m2] − − Both of the terms in the last integral are infrared divergent. The first one is zero in the limit m 0, while the second one gives an infrared divergence that is still → present in the =4 theory, i.e. with m = 0. More precisely putting N 2 2 b I a J A˜ (p)+ C˜(p)= γg N(N 1)δ δ Φ †(p,θ, θ)Φ ( p,θ, θ) [I (p)+ I (p)] , 2 − a J I b − 1 2 h i one has d4k (p k)2 1 I (p)= − = 1 (2π)4 k4 [(p k)2 + m2] − k4 Z − d4k m2 = = (2π)4 k4 [(p k)2 + m2] Z − 1 d4k 1 =2m2 dζ ζ = 4 2 2 2 3 0 (2π) [k +(p ζ + m )(1 ζ)] Z Z − 2m2 1 1 = dζ ζ = 2(4π)2 (p2ζ + m2)(1 ζ) Z0 − m2 1 m2 p2 + m2 = log ǫ + log −(4π)2 (p2 + m2) p2(p2 + m2) m2 where a standard Feynman parameterization has been used. Analogously
d4k p2 I (p)= = 2 (2π)4 k4 [(p k)2 + m2] Z − 1 d4k 1 =2p2 dζ ζ = 4 2 2 2 3 0 (2π) [k +(p ζ + m )(1 ζ)] Z Z − p2 1 1 = dζ ζ = (4π)2 (p2ζ + m2)(1 ζ) Z0 − 1 p2 m2 p2 + m2 = log ǫ + log . −(4π)2 (p2 + m2) (p2 + m2) m2 In the above expressions an infrared regulator ǫ has been introduced. Notice that the total correction is exactly zero on-shell, i.e. for p2=m2. 86 Chapter 3. Perturbative analysis of =4 SYM theory N
To summarize the results, the propagator of the chiral superfields of the =4 N super Yang–Mills theory in =1 superspace is logarithmically infrared-divergent for N any choice of the gauge parameter α = 1. This divergence corresponds to a wave 6 function renormalization for the superfields ΦI and vanishes on-shell. In the Fermi– Feynman gauge α=1 the one-loop correction exactly vanishes.
3.2.2 Propagator of the vector superfield
The one-loop calculation of the propagator of the vector superfield is much more complicated, because many more diagrams are involved producing a large number of contributions. The final result is however completely analogous: in the Fermi– Feynman gauge the one-loop correction is zero, whereas infrared divergences arise for α = 1. In the presence of a mass term for the (anti) chiral superfields no new 6 divergences are generated.
Form a calculational viewpoint the new feature with respect to the ΦΦ† case h i is that there are also diagrams involving the ghosts. There are two multiplets of
ghosts, described by the chiral superfields C and C′. The free propagators for these superfields will be denoted by
a b a b a 1 C ′(z)C (z′) = δ δ (z z′) −→ h ifree b 2 8 −
a b a b a 1 C (z)C ′(z′) = δ δ (z z′) −→ h ifree b 2 8 −
The ghosts are treated exactly like ordinary chiral superfields with the only differ-
ence that there is a minus sign associated with the loops, because C and C′ are anticommuting fields [23].
The corrections to the VV propagator at the one-loop level correspond to the h i diagrams 3.2 Propagators in the =1 formulation 87 N
ΦΦ
a V (z) V (z′) A(z; z′) b −→
Φ†Φ†
ΦΦ†
a V (z) V (z′) B(z; z′) b −→
Φ†Φ
ΦΦ†
a V (z) V (z′) C(z; z′) b −→
CC′
a V (z) V (z′) D (z; z′) b −→ 1
C′C
C′C
a V (z) V (z′) D (z; z′) b −→ 2
CC′
C′C
a V (z) V (z′) D (z; z′) b −→ 3
CC′ 88 Chapter 3. Perturbative analysis of =4 SYM theory N
C′C
a V (z) V (z′) E (z; z′) b −→ 1
CC′
a V (z) V (z′) E (z; z′) b −→ 2
VV
a V (z) V (z′) F (z; z′) b −→ VV
VV
a V (z) V (z′) G(z; z′) b −→
The contributions A to E are rather straightforward to evaluate much in the same
way as the diagrams entering the ΦΦ† propagator. The last two graphs are more h i involved because the free propagator for the V superfield is more complicated for generic values of the parameter α. Moreover the cubic and quartic vertices
i 2 α a b c gfabc D (D V ) V (DαV ) −16√2 1 h i 2 g2f ef V a DαV b D V c D V d , −128 ab ecd α h i
lead to various terms corresponding to the ways the covariant derivatives can act on 3.2 Propagators in the =1 formulation 89 N the V lines. The V 3 vertex in particular
2 b
3 c
1 a gives rise to six different terms. Schematically the calculation goes as follows. A˜(p) is a contribution that appears because of the addition of the mass terms (3.19); it is not present in the =4 the- N ory, i.e. when m=0. It is useful to discuss separately the corrections coming from diagrams A˜, B˜ and C˜ and those obtained from D˜ i, E˜i, F˜ and G˜, since the latter correspond to the one-loop contribution to the vector superfield propagator in the =1 supersymmetric Yang–Mills theory. N The first diagram, A˜, is logarithmically divergent and reads
d4k m2 V a(p,θ, θ)V b( p,θ, θ) A˜(p)=3g2N(N 2 1)δ d4θ − . (3.28) − ab (2π)4 (k2 + m2)[(p k)2 + m2] Z ( − )
For the diagram B˜ application of the Feynman rules gives rise to three different contributions, one quadratically divergent and two logarithmically divergent
d4k 1 B˜(p)=3g2N(N 2 1)δ d4θ − ab (2π)4 (k2 + m2)[(p k)2 + m2] · Z − i α˙ k2V a(p,θ, θ)V b( p,θ, θ) p σµ V a(p,θ, θ) (D Dα)V b( p,θ, θ) + · − − 4 µ αα˙ − h i 1 2 + V a(p,θ, θ) (D D2)V b( p,θ, θ) . (3.29) 4 − h i The tadpole diagram C˜ gives a quadratically divergent result of the form
d4k 1 C˜ = 3g2N(N 2 1)δab d4θ V a(p,θ, θ)V b( p,θ, θ) . (3.30) − − (2π)4 (k2 + m2) − Z Summing and subtracting m2 in the numerator of the first term in (3.29) and putting the three corrections A˜, B˜ and C˜ together gives a net result that is only logarithmi- 90 Chapter 3. Perturbative analysis of =4 SYM theory N
cally divergent d4k 1 A˜(p)+ B˜(p)+ C˜(p) = 3g2N(N 2 1)δ d4θ − − ab (2π)4 (k2 + m2)[(p k)2 + m2] · Z − i α˙ p σµ V a(p,θ, θ) (D Dα)V b( p,θ, θ) + · 4 µ αα˙ − h i 1 2 + V a(p,θ, θ) (D D2)V b( p,θ, θ) . (3.31) 16 − h i Notice that in particular the logarithmically divergent contribution proportional to m2 exactly cancels out. This is crucial because this correction would correspond to a mass renormalization for the vector superfield that is known to be excluded in any gauge theory as well as in supersymmetric theories in general.
The diagrams D˜ i and E˜i are completely analogous to the previous ones, with the only difference that the mass does not appear in the denominators and there is a minus sign associated with the loops. Their sum is logarithmically divergent and takes the form 1 D˜ (p)+ D˜ (p)+ D˜ (p)+ E˜ (p)+ E˜ (p)= g2N(N 2 1)δ 1 2 3 1 2 8 − ab · 4 d k 1 α˙ d4θ ip σµ V a(p,θ, θ) (D Dα)V b( p,θ, θ) + · (2π)4 k2(p k)2 µ αα˙ − Z − n h i 1 2 + V a(p,θ, θ) (D D2)V b( p,θ, θ) . (3.32) 8 − h i All of these corrections do not depend on the gauge parameter γ=α 1 and must − be summed to those coming from the last two diagrams. F˜ exactly vanishes for any α, so that only E˜ needs to be considered. This diagram produces in principle 72 corrections because the Feynman rules give rise to 18 terms (distributing the covariant derivatives associated with the two vertices), each of which splits into 4, since the free propagator itself contains two terms. Many of these contributions can be easily shown to vanish using the properties of the covariant derivatives. In particular one uses
2 2 2 D DαD D =0 ,
3 which follows from D = 0 and use of the (anti) commutation relations for the D’s. It is useful to separate in the non-vanishing part terms proportional to γ and γ2, from the γ-independent terms. The latter combine to give a logarithmically divergent correction that together with that of equation (3.32) cancel the correction 3.2 Propagators in the =1 formulation 91 N
(3.31) coming from the sum A˜ + B˜ + C˜. Actually if m = 0 this sum is finite and 6 exactly vanishes at m=0. As a result the only non-vanishing corrections to the vector superfield propagator at the one loop level come from terms proportional to γ and to γ2 in E˜(p). The former contain an infrared divergent part of the form
µ 4 ν d k σαα˙ σ ˙ pµpν α˙ β˙ J (1)(p)= c γg2δa d4θ ββ V (p,θ, θ) DαD D DβV b( p,θ, θ) , 1 b (2π)4 k4(p k)2 a − Z − n h io but is ultraviolet finite. Furthermore there is a correction, finite both in the ultraviolet and in the infrared regions, proportional to γ2, that reads d4k (σνσµσλ) (p k) k p J (2)(p) = c γ2g2δa d4θ αα˙ − µ ν λ 2 b (2π)4 k4(p k)4 · Z 2 − α˙ (D Dα)V (p,θ, θ) (D2D )V b( p,θ, θ) . · a − nh i h io (1) (2) In the previous expressions for J and J c1 and c2 are non-vanishing numerical constants proportional to N(N 2 1). − In conclusion in the =4 theory, i.e. when m=0, the one-loop correction to the N vector superfield propagator, just like that of the chiral superfield, is infrared singular unless the Fermi–Feynman gauge, α=1, is chosen, in which case it vanishes. Like in the case of the chiral superfield propagator the non-vanishing γ-dependent corrections correspond to a wave function renormalization for the superfield V and are zero on- shell, i.e. for p2=0. The proof of the finiteness of the theory in the presence of the mass terms (3.19) given in [76] is based on naive power counting, which gives the superficial degree of divergence, d , of a diagram in =1 superspace [24] s N d =2 E C , s − − where E is the number of external (anti) chiral lines and C the number of ΦΦ or
Φ†Φ† propagators. In [76] it is also argued that for corrections to the effective action involving only V superfields the requirement of gauge invariance reduces the degree of divergence to
d = C . s − However for the purposes of [77, 78] it appears crucial that no divergences, not even corresponding to a wave function renormalization, are present in total n-point func- tions for any n. The computation of the two-point functions in this section has shown 92 Chapter 3. Perturbative analysis of =4 SYM theory N
that this actually the case at one loop. An argument for the generalization of this result to Green functions with an arbitrary number of external V lines will now be briefly discussed. First note that diagrams involving only vector and ghost super- fields are not modified by the inclusion of (3.19), so that one must only consider graphs containing internal chiral lines. The ultraviolet properties of diagrams that are only logarithmically divergent in the original =4 theory are not modified by the N presence of the mass in the propagators. Thus the contributions to be analyzed are the quadratically divergent ones, that can acquire subleading logarithmic singulari-
ties, plus eventually new diagrams involving ΦΦ and Φ†Φ† propagators. The relevant quadratic divergences come from tadpole diagrams
V˜ an V˜ bn
V˜ a2 V˜ b2
V˜ a1 V˜ b1
The only new diagram containing ΦΦ and Φ†Φ† propagators that must be considered is
V˜ an V˜ bn
V˜ a2 V˜ b2
V˜ a1 V˜ b1
Since all the vertices entering these graphs come from the expansion of the term V V tr e− Φ†e Φ in the action, the only covariant derivatives that are present come from the functional derivatives and must act on the internal (anti) chiral lines to give a non-vanishing result. As a consequence the loop integral in the above diagrams is exactly the same as for the C˜ and A˜ corrections to the VV propagator. Hence 3.2 Propagators in the =1 formulation 93 N summing to the previous diagrams the contribution of
V˜ an V˜ bn
V˜ a2 V˜ b2
V˜ a1 V˜ b1 leads to a net logarithmically divergent correction that is exactly the same as the one obtained in the original =4 theory. Like in the latter case this divergence will N be cancelled by the contributions coming from the other one-loop diagrams involving V and ghost internal lines. The argument given here reinforces the results of [76], at least at the one-loop level, and supports the proposal, put forward in [77, 78], according to which the mass deformed =4 model can be considered a consistent N supersymmetry-preserving regularization for a class of =1 theories. N Notice that for the previous discussion it is not necessary to consider equal masses for all the (anti) chiral superfields. The same results can be proved giving different masses to the three superfields. This can be easily understood since in each diagram considered in this section only one chiral/anti-chiral pair is involved, because the propagators are diagonal in ‘flavour’ space and the vertices containing vector and I (anti) chiral superfields couple ΦI† and Φ with the same index I. Basically this means that the above discussed cancellations apply separately to the contributions of each (anti) chiral superfield. From the viewpoint of the dimensional analysis of [76] having different masses mI is irrelevant. As a consequence one can in particular give mass to only two of the (anti) chiral multiplets. This suggests the possibility of generalizing the approach of [77, 78] to the case of =2 super Yang–Mills theories. A discussion of the effect of a =2 mass N N term in =4 supersymmetric Yang–Mills theory can be found in [81]. N 3.2.3 Discussion
The infrared divergences found in the calculation of the propagators of the chiral and the vector superfields are due to the fact that the vector superfield is dimensionless, so that it contains in particular, as its lowest component, the scalar C that is itself 94 Chapter 3. Perturbative analysis of =4 SYM theory N
dimensionless and hence has a propagator which behaves, in momentum space, like 1 (CC)(k) . h i ∼ k4 The contribution of the scalar C to the VV propagator leads to an infrared diver- h i gence whenever a diagram contains a loop involving a VV line. In the Fermi–Feynman gauge the problem is not present because the choice α = 1 gives a kinetic matrix of the form of equation (3.9) in the component expansion. The corresponding inverse 1 matrix M − in (3.10) shows that no CC free propagator is present. On the con- h i 1 trary any choice α = 1 produces such a propagator, i.e. gives a matrix M − with 6 1 a non-vanishing (M − )11 entry, leading to the previously discussed problems. An explicit calculation in components with α = 1 and without fixing the Wess–Zumino 6 gauge should show problems with infrared divergences analogous to those encountered here. These problems with infrared divergences are not peculiar of the =4 super N Yang–Mills theory, but appear in any supersymmetric gauge theory. Analogous in- frared divergences in Green functions have been observed in [80, 82]. The conclusion proposed in these papers is that there exist no way to remove the infrared diver- gences, so that the choice α=1 is somehow necessary, at least for the computation of gauge-dependent quantities. There are two general theorems concerning infrared divergences in quantum field theory. The first is the Kinoshita–Lee–Nauenberg theorem [83], which deals with infrared divergences in cross sections. It states that no infrared problem is present in physical cross sections of a renormalizable quantum field theory, if the appropriate sums over degenerate initial and final states, associated with soft and collinear parti- cles, are considered. The second theorem was proved by Kinoshita, Poggio and Quinn [84] and concerns Green functions. The statement of the theorem is that the proper (one-particle irreducible) Euclidean Green functions with non-exceptional external momenta are free of infrared divergences in a renormalizable quantum field theory. A
set of momenta pi, i =1, 2,...,n is said to be exceptional if any of the partial sums
p , with I a subset of 1, 2,...,n , i { } i I X∈ vanishes. The reason for the absence of divergences is that the external momenta, if non-exceptional, play the rˆole of an infrared regulator in the Green functions. The proof of the theorem is based on dimensional analysis which does not work in 1 the presence of a k4 propagator. In particular it fails in the case of supersymmetric 3.2 Propagators in the =1 formulation 95 N gauge theories in general gauges, because the adimensionality of the V superfield 1 implies that factors of k4 can enter the loop integrals. The apparent violation of the Kinoshita–Poggio–Quinn theorem can be understood from the viewpoint of the com- ponent formulation: supersymmetric gauge theories, being non-polynomial and thus containing an infinite number of interaction terms, are not formally renormalizable. The choice of the Wess–Zumino gauge makes the theory polynomial. In fact in this case no infrared divergence is found in Green functions and the general theorems are satisfied. However in the case of the =4 theory the choice of the WZ gauge results N in a change of the ultraviolet properties of the model, at least for what concerns gauge dependent quantities, e.g. the propagators. In theories with less supersymmetry, in which ultraviolet divergences are present the problem might seem less severe, since the subtraction of the ultraviolet infinities could cure also the infrared singularity. The infrared problems discussed here are however gauge artifacts and cannot af- fect gauge invariant quantities. The mechanism leading to the cancellation of infrared divergences in gauge independent Green functions has not been verified in detail yet, but can be understood starting from the super Feynman rules. When the propaga- tor VV is inserted into a Feynman graph it is connected to vertices which carry h i covariant derivatives. These derivatives come directly from the form of the action for vertices involving only V fields and from the definition of the functional derivatives for vertices involving (anti) chiral superfields. The situation in diagrams for gauge in- variant quantities is such that at least one covariant derivative can always be brought to act on the V propagator by integrations by parts. In this way an additional factor of the momentum k is generated in the numerator. More precisely for the infrared 1 2 2 2 2 problematic term (D D + D D )δ (θ θ′) one gets − k4 4 − µ 1 2 2 2 2 kµσαα˙ α˙ 2 D (D D + D D )δ (θ θ′) = 4i D D α −k4 4 − − k4 and
µαα˙ α˙ 1 2 2 2 2 kµσ 2 D (D D + D D )δ (θ θ′) =4i D D . −k4 4 − k4 α A rigorous and general check of this mechanism in gauge invariant Green functions has not been achieved yet. 96 Chapter 3. Perturbative analysis of =4 SYM theory N
3.3 Three- and four-point functions of =1 super- N fields
The computation of Green functions with three and four external legs will now be considered. From now on the Fermi–Feynman gauge will be fixed. The three-point functions should in principle suffer from infrared problems of the kind encountered in the preceding section. This claim will not be studied here since the calculation with α = 1 is awkward, even at the one-loop level, because of the huge number of 6 contributions. Four-point functions on the contrary should be infrared finite, because they are directly related to physical scattering amplitudes. Notice, however, that in- frared divergences in the Green function with four external V lines were found in [82]. In that paper beyond the infrared singularity, it was shown that the adimensional- ity of the superfield V implies that it requires a non-linear renormalization, i.e. the
renormalized field VR is a non-linear function of the bare field V
VR = f(V ) .
An example of three-point function will be briefly discussed here and then the more interesting case of a four-point function will be studied in greater detail.
3.3.1 Three-point functions
The simplest three point function to consider corresponds to the correction to the I J K IJK vertex εIJKtr Φ Φ Φ (or ε tr ΦI† ΦJ† ΦK† ), which is determined by one single diagram at the one loop level. However as a less trivial example of computation of three-point functions the one-loop correction to the vertex
J Φb (z)
a c I V (z) Φ †V Φ (z) c −→ I b
a ΦI†(z)
will be considered here. The correction at the one-loop level comes from the following diagrams (the notation for the propagators is the same as in the previous section) 3.3 Three- and four-point functions 97
˜ J Φb (q)
V˜ ( p q) A˜(p; q) c − − −→
˜ a ΦI†(p)
˜ J Φb (q)
V˜ ( p q) B˜(p; q) c − − −→
˜ a ΦI†(p)
˜ J Φb (q)
V˜ ( p q) C˜(p; q) c − − −→
˜ a ΦI†(p)
˜ J Φb (q)
V˜ ( p q) D˜(p; q) c − − −→
˜ a ΦI†(p)
˜ J Φb (q)
V˜ ( p q) E˜(p; q) c − − −→
˜ a ΦI†(p) 98 Chapter 3. Perturbative analysis of =4 SYM theory N
˜ J Φb (q)
V˜ ( p q) F˜(p; q) c − − −→
˜ a ΦI†(p)
In the presence of a mass term for the (anti) chiral superfields the expressions given above are modified by the presence of the mass in the free propagators and further- more there are two additional sets of contributions corresponding to the diagrams
˜ J Φb (q)
V˜ ( p q) G˜(p; q) c − − −→
˜ a ΦI†(p) ˜ J Φb (q)
V˜ ( p q) H˜ (p; q) c − − −→
˜ a ΦI†(p)
The calculation of these diagrams is completely analogous to those that were pre- sented in the previous section. All of these proper contributions can be derived using the Feynman rules. The last two diagrams are finite and vanish in the =4 theory, N i.e. at m=0, as they are proportional to m2; they will not be considered here. The other contributions will be briefly studied in the limit m 0. → The diagram D˜(p, q) is zero as a consequence of the contraction among the colour indices, so that there are five contributions to be calculated. The dimensional analysis of section 1.7 gives a vanishing superficial degree of divergence, corresponding to a logarithmic divergence, for the Green function under consideration. Each diagram actually contains a logarithmically divergent term plus finite terms. A straightforward but rather lengthy computation, based on elementary D-algebra, allows to prove that 3.3 Three- and four-point functions 99
the divergent part of all the diagrams is of the form
4 I abc d k 4 1 J I = c δ f d θ Φ† (p,θ, θ)V ( p q,θ, θ)Φ (q,θ, θ) , log J (2π)4 k2(p k)2 aI b − − c Z − n o where c is a constant. Moreover one finds, from each diagram a finite contribution of the form d4k 1 I = δI f abc d4θ finite J (2π)4 k2(q + k)2(p k)2 Z − α 2 J Φ† (p,θ, θ) c D D D V ( p q,θ, θ) Φ (q,θ, θ) + aI 1 α b − − c µ n h α˙ i +c σ (q + k) DαD V ( p q,θ, θ) ΦJ (q,θ, θ) , 2 αα˙ µ b − − c h i o where c1 and c2 are numerical constants. The sum of all the logarithmically divergent terms contained in the diagrams A˜, B˜, C˜, E˜ and F˜ vanishes. The residual finite part can be shown to be zero as well. In conclusion the one-loop correction to the three- point function Φ†V Φ exactly vanishes in the Fermi–Feynman gauge. The same h i result can be shown to hold for the other three-point functions. The superfield formalism does not lead to significant simplifications in the calcu- lation of three-point functions with respect to the same computation in components (in the Wess–Zumino gauge!); actually for the Green function considered here the number of diagrams to be evaluated is approximately the same in the component formulation. However the power of superspace techniques becomes clear in the com- putation of four-point functions that will be considered in next subsection.
3.3.2 Four-point functions
The computation of four-point functions in components is awkward even at the one- loop level and in the Wess–Zumino gauge. In this case the choice of the WZ gauge should not lead to divergences because the complete four-point function must finally give the physical scattering amplitude. In the =1 superfield formulation the calcu- N lation of four-point functions, though rather lengthy, is much more simple. In this section the computation of the one-particle irreducible one-loop correc- tion to the Green function Φ†ΦΦ†Φ in the Fermi–Feynman gauge will be presented. h i There are several diagrams to be considered: each of them is free of ultraviolet di- vergences, as immediately follows from dimensional analysis. Moreover in the Fermi– Feynman gauge the single corrections are infrared safe. In conclusion one finds a finite and non-vanishing result. 100 Chapter 3. Perturbative analysis of =4 SYM theory N
The first subset of diagrams corresponds to
˜ J ˜ c Φb ΦK†
A˜(p; q; r) −→
˜ a ˜ L ΦI† Φd
together with those obtained from this one by crossing. There are a total of four dia- grams of this kind. Application of the Feynman rules and steps completely analogous to those entering the evaluation of the propagators allow to find the following results. 1 2 A˜ (p,q,r)= g4 κN 2(N 2 1)(δbδd + δ δbd)δI δKI(A)acJL(p,q,r) 1 4 − a c ac J L 1 bdIK where d4k 1 I(A)acJL(p,q,r) = d4θ 1 bdIK (2π)4 k2(p k)2(q + k)2(q + k r)2 · Z − − 2 2 a c L J a (D D )Φ †(p,θ, θ) Φ †(p + q r, θ, θ)Φ (r, θ, θ)Φ (q,θ, θ) + Φ †(p,θ, θ) · I K − d b I · nh 2 2 c i L J 2 a (D D )Φ †(p + q r, θ, θ) Φ (r, θ, θ)Φ (q,θ, θ)+2 (D D )Φ †(p,θ, θ) · K − d b α˙ I · h α˙ c L i J h a i D Φ †(p + q r, θ, θ) Φ (r, θ, θ)Φ (q,θ, θ)+2 D Φ †(p,θ, θ) · − d b α˙ I · h 2 α˙ c i L J h α ia (D D )Φ †(p + q r, θ, θ) Φ (r, θ, θ)Φ (q,θ, θ)+4 (D D )Φ †(p,θ, θ) · − d b α˙ I · h α˙ c i L J h i D D Φ †(p + q r, θ, θ) Φ (r, θ, θ)Φ (q,θ, θ) · α − d b ≡ h d4k i 1 o d4θ KacJL(p,q,r; θ; θ) (3.33) ≡ (2π)4 k2(p k)2(q + k)2(q + k r)2 bdIK Z − − and the constant κ is a group theory factor defined by f f bfgf f dhe = κN 2(N 2 1)(δbδd + δdδb) ; aef cgh − a c a c 1 2 A˜ (p,q,r)= g4 κN 2(N 2 1)(δdδb + δ δbd)δI δKI(A)acJL(p,q,r) , 2 4 − a c ac L J 2 bdIK where d4k 1 I(A)acJL(p,q,r) = d4θ 2 bdIK (2π)4 k2(p k)2(p k r)2(q + k)2(p + q k r)2 · Z − − − − − KacJL(p,q,r; θ; θ); (3.34) · bdIK 3.3 Three- and four-point functions 101
1 2 A˜ (p,q,r)= g4 κN 2(N 2 1)(δbδd + δdδb)δI δKI(A)acJL(p,q,r) , 3 4 − a c a c J L 3 bdIK where d4k 1 I(A)acJL(p,q,r) = d4θ 3 bdIK (2π)4 k2(p k)2(q + k)2(q + k r)2 · Z − − 2 a 2 J c L D Φ †(p,θ, θ) D Φ (q,θ, θ) Φ †(p + q r, θ, θ)Φ (r, θ, θ)+ · I b K − d nh µ i α˙ a α J c +8iσ (p k r) D Φ †(p,θ,θ) D Φ (q,θ, θ) Φ †(p + q r, θ, θ) αα˙ − − µ I b K − · L 2 a J c Φ (r, θ, θ) 16(p hk r) Φ †(p,θ,i θ)Φ (q,θ, θ)Φ†(p + q r, θ, θ) · d − − − I b K − · ΦL(r, θ, θ) ; (3.35) · d
1 2 A˜ (p,q,r)= g4 κN 2(N 2 1)(δbδd + δdδb)δI δKI(A)acJL(p,q,r) , 4 4 − a c a c J L 4 bdIK where d4k 1 I(A)acJL(p,q,r) = d4θ 4 bdIK (2π)4 k2(p k)2(p k r)2(q + k)2 · Z − − − 2 a J c 2 L D Φ †(p,θ, θ) Φ (q,θ, θ)Φ †(p + q r, θ, θ) D Φ (r, θ, θ) + · I b K − d nh µ i α˙ a J c +8iσ (q + k) D Φ †(p,θ, θ) Φ (q,θ, θ)Φ †(p+ q r, θ, θ) αα˙ µ I b K − · α L 2 a J c D Φ (r, θ, θ) h 16(q + k) Φi†(p,θ, θ)Φ (q,θ, θ)Φ †(p + q r, θ, θ) · d − I b K − · Φ L(r, θ, θ) . (3.36) · d The second kind of contribution corresponds to the diagram
˜ J ˜ c Φb ΦK†
B˜(p; q; r) −→
˜ a ˜ L ΦI† Φd
In this case the diagrams obtained by crossing are identical to the one depicted, so that they are accounted for by giving the correct weight to B˜(p,q,r). For this contribution one finds g4 B˜(p,q,r)= κN 2(N 2 1)(δbδd + δdδb)(δI δK + δI δK)I(B)acJL(p,q,r) , 6 − a c a c J L L J bdIK 102 Chapter 3. Perturbative analysis of =4 SYM theory N
where d4k 1 I(B)acJL(p,q,r) = d4θ bdIK (2π)4 k2(p k)2(p + q r)2(q + k)2 · Z − − 2 a J c 2 L D Φ †(p,θ, θ) Φ (q,θ, θ)Φ †(p + q r, θ, θ) D Φ (r, θ, θ) + · I b K − d nh µ i α˙ a J c +8iσ (p k) D Φ †(p,θ, θ) Φ (q,θ, θ)Φ †(p+ q r, θ, θ) αα˙ − µ I b K − · α L 2 a J c D Φ (r, θ, θ) h 16(p k) Φi†(p,θ, θ)Φ (q,θ, θ)Φ †(p + q r, θ, θ) · d − − I b K − · ΦL(r, θ, θ) . (3.37) · d
The next one-loop correction is
˜ J ˜ c Φb ΦK†
C˜(p; q; r) −→
˜ a ˜ L ΦI† Φd
This contribution is trivially zero because it contains the product
δ (θ θ )δ (θ θ ) 0 . 4 1 − 2 4 1 − 2 ≡ This would not be true with a choice of gauge different from the Fermi–Feynman one, i.e. with α = 1, because in that case there would be projectors acting on the δ’s. 6 The vanishing of C˜(p,q,r), that is completely trivial in the superfield formulation, corresponds, in the component formulation, to a complicated cancellation among various terms coming from graphs with the same topology. Another subset of diagrams contains
˜ c ˜ J ΦK† Φb
D˜(p; q; r) −→
˜ a ˜ L ΦI† Φd 3.3 Three- and four-point functions 103
plus the crossed versions of this one. There are three inequivalent crossed diagrams. The result of the calculation is the following.
1 2 D˜ (p,q,r)= g4 κN 2(N 2 1)(δ δbd + δdδb)(δI δK δI δK)I(D)acJL(p,q,r) , 1 4 − ac a c J L − L J 1 bdIK where d4k 1 I(D)acJL(p,q,r) = d4θ 1 bdIK (2π)4 k2(p k)2(k + r p q)2(p k r)2 · Z − − − − − 2 a 2 J c L D Φ †(p,θ, θ) D Φ (q,θ, θ) Φ †(p + q r, θ, θ)Φ (r, θ, θ)+ · I b K − d nh µ i α˙ a α J c +8iσ (k + r p) D Φ †(p,θ,θ) D Φ (q,θ, θ) Φ †(p + q r, θ, θ) αα˙ − µ I b K − · L 2 a J c Φ (r, θ, θ) 16(k +h r p) Φ †(p,θ,i θ)Φ (q,θ, θ)Φ†(p + q r, θ, θ) · d − − I b K − · ΦL(r, θ, θ) ; (3.38) · d
1 2 D˜ (p,q,r)= g4 κN 2(N 2 1)(δ δbd + δbδd)(δI δK δI δK)I(D)acJL(p,q,r) , 2 4 − ac a c L J − J L 2 bdIK where d4k 1 I(D)acJL(p,q,r) = d4θ 2 bdIK (2π)4 k2(p k)2(k + r p q)2(p + q k)2 · Z − − − − 2 a J c 2 L D Φ †(p,θ, θ) Φ (q,θ, θ)Φ †(p + q r, θ, θ) D Φ (r, θ, θ) + · I b K − d nh µ i α˙ a J c +8iσ (k p q) D Φ †(p,θ, θ) Φ (q,θ, θ)Φ †(p + q r, θ, θ) αα˙ − − µ I b K − · α L 2 a J c D Φ (r, θ, θ) 16(h k p q) Φi†(p,θ, θ)Φ (q,θ, θ)Φ †(p + q r, θ, θ) · d − − − I b K − · Φ L(r, θ, θ) ; (3.39) · d
1 2 D˜ (p,q,r)= g4 κN 2(N 2 1)(δdδb + δ δbd)(δI δK δI δK)I(D)acJL(p,q,r) , 3 4 − a c ac J L − L J 3 bdIK where d4k 1 I(D)acJL(p,q,r) = d4θ 3 bdIK (2π)4 k2(p k)2(k + r p q)2(p + q k)2 · Z − − − − a J 2 c 2 L Φ †(p,θ, θ)Φ (q,θ, θ) D Φ †(p + q r, θ, θ) D Φ (r, θ, θ) + · I b K − d n µ a h J α˙ ic 8iσ (p k r) Φ †(p,θ, θ)Φ (q,θ, θ) D Φ †(p + q r, θ, θ) − αα˙ − − µ I b K − · α L 2 a J c D Φ (r, θ, θ) 16(p k r) Φ †(p,θ,hθ)Φ (q,θ, θ)Φ †(p + qi r, θ, θ) · d − − − I b K − · ΦL(r, θ, θ) ; (3.40) · d
104 Chapter 3. Perturbative analysis of =4 SYM theory N
1 2 D˜ (p,q,r)= g4 κN 2(N 2 1)(δbδd + δ δbd)(δI δK δI δK)I(D)acJL(p,q,r) , 4 4 − a c ac L J − J L 4 bdIK where d4k 1 I(D)acJL(p,q,r) = d4θ 4 bdIK (2π)4 k2(p k)2(k + r p q)2(p + q k)2 · Z − − − − a 2 J 2 c L Φ †(p,θ, θ) D Φ (q,θ, θ) D Φ †(p + q r, θ, θ) Φ (r, θ, θ)+ · I b K − d n µ a h α J α˙ ic 8iσ (p + q k) Φ †(p,θ,θ) D Φ (q,θ, θ) D Φ †(p + q r, θ, θ) − αα˙ − µ I b K − · L 2 a J c Φ (r, θ, θ) 16(p + q k) Φ †(p,θ, θ)Φ (q,θ, hθ)Φ †(p + q r, θ, θ) i · d − − I b K − · ΦL(r, θ, θ) . (3.41) · d The last one-loop family of diagrams is formed by the one below plus those ob- tained by crossing.
˜ J ˜ c Φb ΦK†
E˜(p; q; r) −→
˜ a ˜ L ΦI† Φd
The resulting contributions to the Green function are 1 2 g4 E˜ (p,q,r)= κN 2(N 2 1)δbδd + δ δbd)δI δK I(E)acJL(p,q,r) , 1 4 2 − a c ac J L 1 bdIK where d4k 1 I(E)acJL(p,q,r) = d4θ 1 bdIK (2π)4 k2(p k)2(p + q k)2 · Z − − a J c L Φ †(p,θ, θ)Φ (q,θ, θ)Φ †(p + q r, θ, θ)Φ (r, θ, θ) ; (3.42) · I b K − d n o 1 2 g4 E˜ (p,q,r)= κN 2(N 2 1)(δdδb + δ δbd)δI δK I(E)acJL(p,q,r) , 2 4 2 − a c ac L J 2 bdIK where d4k 1 I(E)acJL(p,q,r) = d4θ 2 bdIK (2π)4 k2(p k)2(r k)2 · Z − − a J c L Φ †(p,θ, θ)Φ (q,θ, θ)Φ †(p + q r, θ, θ)Φ (r, θ, θ) ; (3.43) · I b K − d n o 3.3 Three- and four-point functions 105
1 2 g4 E˜ (p,q,r)= κN 2(N 2 1)(δbδd + δ δbd)δI δKI(E)acJL(p,q,r) , 3 4 2 − a c ac L J 3 bdIK where d4k 1 I(E)acJL(p,q,r) = d4θ 3 bdIK (2π)4 k2(p + q k)2(k r)2 · Z − − a J c L Φ †(p,θ, θ)Φ (q,θ, θ)Φ †(p + q r, θ, θ)Φ (r, θ, θ) ; (3.44) · I b K − d n o 1 2 g4 E˜ (p,q,r)= κN 2(N 2 1)(δdδb + δ δbd)δI δK I(E)acJL(p,q,r) , 4 4 2 − a c ac L J 4 bdIK where d4k 1 I(E)acJL(p,q,r) = d4θ 4 bdIK (2π)4 k2(q + k)2(k + r p)2 · Z − a J c L Φ †(p,θ, θ)Φ (q,θ, θ)Φ †(p + q r, θ, θ)Φ (r, θ, θ) . (3.45) · I b K − d The sum of alln the preceding terms results in a finite and non vaonishing total one-loop correction to the four point function. The final expression contains terms with six different tensorial structures 2 6 1 4 2 2 (i) Φ†ΦΦ†Φ = κ g N (N 1) G , h i 4 − i=1 X where 2 G(1) = δbδdδI δK I(A)acJL + I(A)acJL + I(B)acJL I(D)acJL+ a c J L 1 bdIK 3 bdIK 3 bdIK − 2 bdIK I(D)acJL + I(E)acJL + I(E)acJL − 4 bdIK 1 bdIK 3 bdIK (A)acJL (D)acJL (D)acJL (D)acJL G(2) = δ δbdδI δK I + I I + I + ac J L 1 bdIK 1 bdIK − 2 bdIK 3 bdIK (D)acJL (E)acJL (E)acJL I + I + I − 4 bdIK 1 bdIK 3 bdIK 2 G(3) = δbδdδI δK I(A)acJL + I(B)acJL + I(D)acJL + I(D)acJL a c L J 2 bdIK 3 bdIK 2 bdIK 4 bdIK 2 G(4) = δdδbδI δK I(A)acJL + I(A)acJL + I(B)acJL I(D)acJL a c L J 2 bdIK 4 bdIK 3 bdIK − 1 bdIK − I(D)acJL + I(E)acJL + I(E)acJL − 3 bdIK 2 bdIK 4 bdIK 2 G(5) = δdδbδI δK I(A)acJL + I(B)acJL + I(D)acJL + I(E)acJL a c J L 3 bdIK 3 bdIK 1 bdIK 3 bdIK G(6) = δ δbdδI δK I(A)acJL I(D)acJL + I(D)acJL I(D)acJL+ ac L J 4 bdIK − 1 bdIK 2 bdIK − 3 bdIK (D)acJL (E)acJL (E)acJL + I4 bdIK + I2 bdIK + I4 bdIK 106 Chapter 3. Perturbative analysis of =4 SYM theory N
In the presence of a mass term for the (anti) chiral superfields the expressions given above are slightly modified by the presence of the mass in the free propagators and furthermore there are two additional sets of contributions corresponding to the diagrams
˜ c ˜ L ΦK† Φd
F˜(p; q; r) −→
˜ a ˜ J ΦI† Φb
˜ J ˜ L Φb Φd
G˜(p; q; r) −→
˜ a ˜ c ΦI† ΦK†
Both of these graphs give corrections proportional to m2, that can be calculated much in the same way as the previous ones.
With a different choice of gauge the computation of four-point Green functions is more complicated. Single diagrams involving vector superfield propagators contain new contributions, some of which are infrared divergent. The correction C˜ is not zero anymore, because there are projection operators acting on the δ-functions. Moreover further diagrams must be included in the calculation at the same order as a conse- quence of the non-vanishing of the one-loop correction to the vertices. For example one must consider diagrams such as 3.3 Three- and four-point functions 107
˜ J ˜ L Φb Φd
˜ a ˜ c ΦI† ΦK†
The techniques illustrated in this chapter for the calculation of Green functions in the =1 formalism can be applied to correlation functions of composite operators N as well. Green functions of gauge invariant composite operators such as those that form the multiplet of currents (equations (2.17) and (2.18)) play a crucial rˆole in the correspondence with type IIB superstring theory on AdS space to be discussed in chapter 5. The application of =1 superspace to this problem will be sketched in N the final chapter. It can be shown that the extension of the formalism to the case of composite operators is rather straightforward, the fundamental difference being that the complete Green functions and not the proper parts must be considered. Chapter 4
Nonperturbative effects in =4 super Yang–Mills theory N
In the last chapter the quantum properties of =4 supersymmetric Yang–Mills the- N ory have been analyzed at the perturbative level, this chapter is devoted to the study of some non-perturbative effects. Instanton calculus will be employed to compute various Green functions of gauge invariant composite operators in the semiclassical approximation. It is well known that in gauge field theories non-perturbative effects can modify in a dramatic way the perturbative structure of the vacuum. Instanton calculus has proved to be an extremely powerful tool in the analysis of non-perturbative phe- nomena in quantum field theories, particularly in the case of supersymmetric gauge theories. The rˆole of instantons in Yang–Mills theories was pointed out in [85] at the classical level and then at the quantum level in a fundamental work by ‘t Hooft [86]. Instantons are non trivial finite action solutions of the classical Euclidean equa- tions of motion of the model. The computation of Green functions in the instanton background allows to study non-perturbative properties of the quantum theory. In non-supersymmetric non Abelian gauge theories the contribution of instantons to Green functions turns out to be divergent in the semiclassical approximation [86]. However in the supersymmetric case explicit instanton calculations can be performed leading to finite and well defined results, thanks to the cancellation between fer- mionic and bosonic quantum fluctuations. Compensations of quantum effects due
108 109
bosons and fermions take place in the instanton background and allow to explicitly compute vacuum expectation values (vev’s) of composite operators (condensates). These cancellations are reminiscent of boson and fermion compensations in internal loops at the perturbative level. The condensates play the rˆole of order parameters in non-perturbative phenomena such as for instance the spontaneous breaking of chiral symmetries or even of supersymmetry in certain models [87]. Furthermore instanton calculus together with holomorphy properties, has allowed to derive some exact re- sults in supersymmetric theories [33] and, more recently, to obtain non trivial checks of the exact result proposed in [34]. As was discussed in previous chapters supersymmetric gauge theories typically possess a moduli space of vacua parameterized by the vev’s of scalar fields. The literature on instanton effects in asymptotic free theories includes examples in which the computations are performed with zero vev’s for all the scalars [87] as well as examples in which non vanishing vev’s for the scalar fields parameterizing the moduli space are considered [88]. The =4 super Yang–Mills theory has a (exactly) vanish- N ing β-function and a moduli space of vacua, =R6k , where k is the rank of the M × Sk gauge group and the group of permutations of k elements, parameterized by the Sk vev’s of six real scalar fields, ϕi . The superconformal phase, corresponding to the h i origin of , i.e. ϕi =0, i, cannot be obtained naively as the limit of the theory M h i ∀ in the Coulomb phase. In this chapter the approach of [87] will be applied to the computation of various condensates corresponding to operators in the supermultiplet of currents (2.17), (2.18). Four-, eight- and sixteen-point functions of bilinear oper- ators belonging to the current supermultiplet will be computed in the semiclassical approximation in the case of a SU(2) gauge group. In particular the exact complete spatial dependence will be determined for a four-point function at lowest order. Due to the vanishing of the β-function and of the associated chiral anomalies, the correlation functions that will be studied, unlike the cases encountered in theories with less supersymmetry, receive a non-vanishing perturbative contribution as well. These Green functions will reconsidered in the final chapter in the context of the correspondence with type IIB superstring theory compactified on AdS S5. It will 5 × be shown that the difficulties encountered in perturbative calculations in chapter 3 do not affect the computations discussed here. The chapter is organized as follows. The general features of instanton calculus in supersymmetric gauge theories are briefly discussed in sections 1, 2 and 3. Sections 4.4, 4.5 and 4.6 report on original results in the case of =4 supersymmetric Yang– N 110 Chapter 4. Non-perturbative effects in =4 SYM theory N
Mills theory, following [89]. The final section describes the generalizations of the results of [89] to the case of a SU(N) gauge group and to the K-instanton sector in the large N limit, that were given respectively in [90] and [91, 92].
4.1 Instanton calculus in non Abelian gauge theo- ries
In this section the semiclassical quantization in the background of an instanton con- figuration will be reviewed for the case of a pure Yang–Mills theory. The extension to fermions will be discussed in the next section. In quantum field theory one is interested in computing vacuum expectation values of operators (φ), that in general are polynomials in the elementary fields φ. In O Euclidean space one must calculate
S[φ] (x ,...,x ) = [ φ]e− ℏ (φ(x ),...,φ(x )) , (4.1) hO 1 n i D O 1 n Z where the dependence on ℏ has been explicitly indicated in order to discuss the semiclassical approximation. The classical vacuum of the theory corresponds to a constant expectation value φ for the elementary fields and a perturbative series is obtained expanding the h i interaction part of the action S[φ] for small fluctuations, δφ, around the classical vacuum. In instanton calculations one considers a situation in which there exists a non-trivial solution of the classical equations of motion with finite Euclidean action. Denoting by φ such a field configuration one can then study quantum fluctuation around φ = φ in the limit ℏ 0. Putting →
φ(x)= φ(x)+ δφ(x)= φ(x)+ ℏ1/2η(x)
and expanding S[φ] in powers of ℏ in the form
ℏ δ2S S[φ]= S[φ]+ d4xd4y η(x)η(y)+ O(ℏ3/2) , 2 δφ(x)δφ(y) Z φ=φ
the vacuum expectation value of an operator [φ] can be evaluated in the semiclas- O 4.1 Instantons in Yang–Mills theories 111
sical limit by a saddle point approximation 1
S[φ+η] (x ,...,x ) = [ η]e− ℏ [(φ + η)(x )] = hO 1 n i D O i Z 1 2 − 2 S δ S = (φ) e− ℏ det (1 + O(ℏ)) . (4.2) O δφ(x)δφ(y) " φ=φ#
Notice that in principle one could consider a classical solution φ with infinite Eu- clidean action, but this would give a vanishing contribution in the semiclassical ap- S[φ]/ℏ proximation because of the factor e− in (4.2).
4.1.1 Yang–Mills instantons
The classification of non-trivial solutions of the classical Euclidean equations of mo- tion for the Yang–Mills system has been given in [85], where an explicit solution has also been constructed. Another (multi-center) solution has been proposed by ’t Hooft, who also studied the semiclassical quantization [86]. The case of a SU(2) gauge group will be discussed first and then the extension to SU(N) will be sketched. The Euclidean Yang–Mills action is
1 4 µν SYM[A]= d x tr(Fµν F ) , (4.3) −4drg2 Z where the gauge field Aµ and the field strength Fµν are respectively
a Aµ = AµTa F µν = ∂µAν ∂ν Aµ + [Aµ, Aν]= ∂ Aa ∂ Aa + if a Ab Ac T − µ µ − ν µ bc µ ν a and T a are generators of the gauge group in the representation r, satisfying [T a, T b]= ab c a b ab if cT and tr T T = δ dr, dr being the Dynkin index of the representation. Gauge transformations take the form
Ω A (x) A (x)=Ω(x) [A (x)+ i∂ ] Ω†(x) , µ −→ µ µ µ
a with Ω(x)= eiαa(x)T . The study of finite Euclidean action configurations is performed much in the same way as for the monopoles of the Georgi–Glashow model in appendix B. The requirement of finite action implies that Aµ must approach asymptotically, i.e. for 1This relation holds for bosonic fields φ, see the next section for the case of fermions. 112 Chapter 4. Non-perturbative effects in =4 SYM theory N
x , a pure gauge configuration. This leads to a topological classification of the | |→∞ field configurations in homotopy classes. The quantity which characterizes the finite energy solutions is the Pontryagin index
1 K[A]= d4xF a F˜µν , (4.4) 32π2 µν a Z ˜ 1 ρσ where F = 2 εµνρσF . The Yang–Mills action satisfies a “Bogomol’nyi bound”
8π2 1 8π2 S [A]= K[A]+ d4x F a F˜a F µν F˜µν K[A] . (4.5) YM ± g2 8π2 µν ∓ µν a ∓ a ≥ g2 | | Z Solutions of the classical equations of motion correspond to local minima of the action functional, so that in each homotopy class they saturate the bound in (4.5). As a consequence they are determined by the condition
F a = F˜a . µν ± µν
In the K=0 sector the minimum is reached on the trivial vacuum, Aµ=0, while the configurations minimizing S[A] for K = 0 are (anti) instantons. Self-dual and anti | | 6 self-dual configurations are referred to as respectively instantons and anti-instantons. In the K=1 sector and for a gauge group SU(2) the explicit solution was found in [85, 86] through the ansatz
σ A = i µν ∂ν log Φ , (4.6) µ 2 where Φ is real scalar field. Substituting into the equations of motion one obtains the equation for Φ
2Φ =0 . Φ A solution can be written in the form ρ2 Φ=1+ . (4.7) (x x )2 − 0
Φ depends on 5 free parameters, ρ associated with the “size” of the instanton and x0 with its position. The corresponding instanton solution is yν A = 2iρ2ηa , (4.8) µ − µν y2(y2 + ρ2) 4.1 Instantons in Yang–Mills theories 113
where y = x x and ηa are the anti self-dual ’t Hooft symbols − 0 µν εaµν for µ, ν =1, 2, 3 ηa = ηa = µν − νµ δaµ for ν =4 . − The general solution depends on 3 additional parameters corresponding to the free- dom of performing global gauge rotations. Actually the Yang–Mills system is classi- cally invariant under the four-dimensional conformal group, which partially overlaps with global gauge transformations. In [93] it has been proved that the most general one-instanton solution depends on 8 parameters, referred to as collective coordinates or moduli, that in the following will be denoted by βi. The moduli space, i.e. the space parameterized by the coordinates β , will be indicated by . The field (4.8) i M has a singularity at y=0, which can be removed by a correspondingly singular gauge transformation, so that one can construct a solution with no singularities at any x
ν a a iθaT a y iθaT A = 2ie η e− , (4.9) µ − µν (y2 + ρ2)
a where ηµν are the self-dual ’t Hooft symbols
εaµν for µ, ν =1, 2, 3 ηa = ηa = µν − νµ δaµ for ν =4 , which satisfy
a σµν = ηµνσa , with σa, a =1, 2, 3 the standard Pauli matrices. The associated field strength is
2 a a iθaT ρ iθaT F = F [A]= 2ie σ e− . (4.10) µν µν − µν (y2 + ρ2)2 The ansatz of (4.7) can be generalized to a n-instanton solution considering the scalar field n ρ2 Φ(x)=1+ i . (i) 2 i=1 (x x0 ) X − (n) (i) Substituting into (4.6) yields a solution, Aµ , with singularities at xµ = x0 µ, which can be made regular by a suitable singular gauge transformation. For such a config- uration one can prove that
(n) (1) S[A ]= nS[A ] , 114 Chapter 4. Non-perturbative effects in =4 SYM theory N
(n) so that Aµ (x) is a solution in the homotopy class K[A]=n. The number of collective coordinates for the n-instanton solution is 8n 3 (with n > 1). There are 8 moduli − for each single instanton, but one must subtract 3 global rotations that are already accounted for by the relative rotations. This result can be proved to hold in general, even if the n instantons are not widely separated, [94].
4.1.2 Semiclassical quantization
To perform a semiclassical quantization around an instanton configuration according
to (4.2) one needs the expansion of S[A] around Aµ = Aµ, obtained putting Aµ =
Aµ + Qµ, where Qµ denotes the quantum fluctuation. A straightforward calculation gives the gauge kinetic operator in the instanton background δ2S[A] M ab = = D2(A)δ + D (A)D (A) 2F δab , µν δAa δAb − µν µ ν − µν µ ν A
where Dµ(A)= ∂µ + i[Aµ, . ] is the covariant derivative. In the instanton background the Faddeev–Popov quantization must be suitably
modified, because the kinetic operator Mµν possesses 8 new zero modes (in the case of a SU(2) gauge group under consideration) beyond those associated with gauge symmetry. These new zero-modes correspond to the symmetries of the action S[A] that are broken by the instanton configuration and are related to the 8 collective
coordinates βi in the general solution (4.9). In fact since Aµ is a solution for any βi
δ2S[A] ∂A d4y ν (y, β)=0 , δAµδAν ∂βi Z Aβ
so that (4.2) is formally divergent. This divergence is completely analogous to those introduced in the partition function of gauge theories by the functional integration along the gauge orbits, that is dealt with by the Faddeev–Popov procedure. The correct way of performing the quantization in the instanton background was discussed in [95]. The general instanton configuration will be parameterized as
Ω Aµ (x, β) = Ω†AµΩ+ iΩ†∂µΩ ,
where Ω denotes an arbitrary gauge transformation and β the set collective coordi- nates, 8 in the case of SU(2) and K[A]=1 and in general 4NK[A] in the K-instanton sector for a SU(N) gauge group. The basic idea is then to constrain `ala Faddeev– Ω Popov the quantum fluctuations Qµ to be orthogonal to Aµ as Ω varies in G and βi 4.1 Instantons in Yang–Mills theories 115
in the moduli space . In the partition function of the system 2 M
S[A] Z = [ A]e− D Z one inserts
Ω Ω Ω δA Ω ∂A 1 = ∆ [ ha(x)] dβ δ (A A , µ ) δ (A A , µ ) , FP D i µ − µ δha µ − µ ∂β G a,x i ! i ! Z Y ZM Y (4.11)
a where Ω(x) is parameterized as Ω(x) = eih (x)Ta and the scalar product in the argu- ments of the δ-functions is 1 (f ,gµ)= d4x tr(f (x)gµ(x)) . (4.12) µ 2 µ Z
Equation (4.11) defines the Faddeev–Popov determinant ∆FP. Some algebraic steps allow write Z in the form
S[A] Z = [ A ] dβ e− ∆ (A, β) D µ i FP · i Z Z Y ∂A δ tr T aD (A) Aµ A (x) δ (A A , µ ) . · µ − µ − µ ∂β i In this relation an infinite constant, coming from the integration over the group algebra, has been dropped, since in the following the above construction will be employed in the computation of normalized vev’s of operators. The Faddeev–Popov determinant in the above expression can be computed much in the same way as in the case of the ordinary of gauge theories. To compute expectation values in the semiclassical approximation one only needs
∆FP(A, β) on the classical solution A. It can be shown that in this case ∆FP can be written as
2 ∆FP(A, β) = det Dab(A)δ(x y) det (dij(A, β)) , (4.13) − i,j A=A x,y a,b
where
(i) µ(j) dij(A, β)=(ˆaµ (β), ˆa (β)) 2From now on ℏ will be set equal 1. 116 Chapter 4. Non-perturbative effects in =4 SYM theory N
The vectorsa ˆ in the above equation are given by a (i) a ∂Aµ (i) a aˆµ (x, β)= + ξµ (x) ∂βi (i) a and ξµ ) are determined by the transversality condition
ab (i) Dµ (A)ˆaµ b(x, β)=0 . In conclusion introducing ghost fields, c and c to rewrite the determinants as Gaussian integrals the expectation value of a gauge invariant operator [A] in the O semiclassical approximation takes the form
2 8π K[A] (x ,...,x ) = e− g2 hO 1 n iA · 1 µ g.f. ν 2 aˆ Q Mµν Q cD (A)c [ Q c c] dβ k k e− 2 − [A] D µD D i √ O i 2π R R Z Y = 1 µ (0)g.f. ν · Q Mµν Q c2c [ Qµ c c]e− 2 − D D D R R Z 1 2 g.f. − 2 2 8π K[A] aˆ det′ Mµν det D (A) − g2 = e dβi k k [A] 1 , (4.14) √ O 2 i 2π (0)g .f. − 2 Z Y det′ Mµν det g.f. (0)g.f. where the ‘gauge-fixed’ kinetic operators Mµν and Mµν are M g.f. = D2(A)δ 2F [A] , M (0)g.f. = 2δ . µν − µν − µν µν − µν 4.1.3 Bosonic zero-modes
To explicitly compute expectation values through equation (4.14) one needs to know (i) the zero-mode vectorsa ˆµ , whose norms enter the integration measure. Since the (i) kinetic operator is modified by the gauge-fixing term the vectorsa ˆµ are given by
(i) ∂Aµ (i) aˆµ (x)= (x, β)+ Dµ(A)Λ (x) , ∂βi for a suitable function Λ(i)(x), such that
aˆ(i)(x) Ker M g.f. . µ ∈ µν To compute the zero-mode vectors it is convenient to refer to the singular instanton solution. For the case of a SU(2) gauge group, after rescaling A 1 A , it reads µ → g µ 2 a a i iθaT ρ (x x0)µ iθaT A (x, β)= e σ − e− , (4.15) µ −g µν (x x )2[(x x )2 + ρ2] − 0 − 0 4.1 Instantons in Yang–Mills theories 117
µ where the 8 moduli are βi =(x0 , ρ, θa).
Zero-modes associated with translations.
In this case β xµ and there are four zero-modes. The transverse vectors are i → 0 ∂ aˆµ(ν) = ν Aµ + Dµ(A)Aν = Fµν [A] . ∂x0 The computation of the norm is straightforward using (4.15), the result is
2√2 π aˆ = . k µ(ν)k g
Zero-modes associated with dilatations.
(dil) The corresponding collective coordinate is the size of the instanton ρ.a ˆµ is simply given by
∂A 2i ρyν aˆ(dil) = µ = σ , µ ∂ρ − g µν (y2 + ρ2)2 so that for the norm one obtains 4π aˆ(dil) = . k µ k g
Zero-modes associated with gauge rotations.
The relevant collective coordinates are the angles θa in (4.15) and one can put
∂A i aˆ(a) = µ + D (A)Λ(a) = D (A) T a +Λ(a) D (A)φ(a) , µ ∂θ µ µ −g ≡ µ a where the scalar field φ(a), defined by the last equality in the above equation, must satisfy i D2(A)φ(a) =0 , lim φ(a) = T (a) . x −g | |→∞ The solution is i r2 φ(a) = T (a) with r = x . −g r2 + ρ2 | | 118 Chapter 4. Non-perturbative effects in =4 SYM theory N
The norms are then 2πρ aˆ(a) = . k µ k g The above eight vectors can be shown to be orthogonal with respect to the scalar product (4.12).
Having determined the vectorsa ˆµ one can thus replace the integration over the zero-modes in (4.14) with an integration over the instanton collective coordinates and write the integration measure as 4 aˆ 1 2√2π 4π 2πρ 3 dµ = dβ k k = d4x dρd3θ = B i √ (2π)4 g g g 0 i 2π ! Y 27π4 = ρ3d4x dρd3θ . (4.16) g8 0 The generalization to larger gauge groups, in particular to SU(N), is rather straightforward [96, 97]. First it can be shown [98], that the mapping of S3 into a generic group G, which determines the topological classification of the instanton solutions, can be continuously deformed to a mapping into SU(2). Hence the classifi- cation of classical solutions into homotopy classes holds in general. Furthermore one can construct a solution in the case of a gauge group G by the embedding of a SU(2) instanton. In particular for G=SU(N) one can consider a configuration 2i ηa xν λ A = µν k , µ g x2(x2 + ρ2) 2
where λk, k =1, 2, 3, are the first three generators of the fundamental representation
of SU(N). Under the SU(2) subgroup of SU(N) (λ1,λ2,λ3) transforms as a triplet, whereas the other generators are organized into 2(N 2) doublets and the remaining − are singlets. In conclusion, in addition to eight zero-modes of the SU(2) instanton, there are 4(N 2) additional zero-modes. The corresponding vectors take the form − [97]
1 λr x2 2 aˆr = D (A) . µ µ g x2 + ρ2 " # 4.2 Instanton calculus in the presence of fermionic fields
In this section the semiclassical quantization in the background of an instanton config- uration will be generalized to the case of a non Abelian gauge theory with interacting 4.2 Instantons and fermions 119
fermions. The Euclidean action of the model is
4 µ S[ψ, ψ, A]= SYM[A]+ SA[ψ, ψ]= SYM[A]+ d x ψ [iDµ(r)] σ ψ , (4.17) Z where SYM is the Yang–Mills action (4.3) and Dµ(r) denotes the covariant derivative in the representation r of the gauge group G (see appendix A for the notations). In (4.17) ψ is a Weyl fermion; an equivalent description can be given in terms of a Dirac spinor Ψ, introducing a fictitious right chirality Weyl spinor and writing the action as
4 SA[Ψ, Ψ] = d x Ψ iD/ L(r) Ψ , Z where
µ 0 σ ∂µ D/ (r)= . L µ σ (∂µ + Aµ) 0 !
4.2.1 Semiclassical approximation
In the full interacting theory one is interested in computing expectation values of fields (x ,... ,x ) that are composite operators made up from the elementary fields ψ, ψ O 1 n and Aµ. In the semiclassical approximation one considers fluctuations, Aµ = Aµ +Qµ, around a classical solution Aµ; for the fermionic part of the action it is sufficient to restrict oneself to SA[ψ, ψ]. In the spirit of the semiclassical approximation, in the computation of expectation values of operators depending on Aµ as well as on fermionic fields, the functional integration over the latter is performed first and then the result is substituted into (4.14). Thus as a starting point one considers
SA[Ψ,Ψ] [Ψ, Ψ] = [ Ψ Ψ]e− [Ψ, Ψ] hO iA D D O Z and evaluates this integral through a saddle point approximation. According to the properties of Berezin integration (see appendix A) the last in- tegral yields a non-vanishing result only if all the integrations over the fermionic variables are correctly saturated. Expanding Ψ and Ψ into eigenfunctions of the Dirac operator
(n) (n) (n) (n) iD/ L(r)f = λnf , iD/ R(r)g = λn∗ g , 120 Chapter 4. Non-perturbative effects in =4 SYM theory N
gives
m (0) (0) (n) Ψ = bi fi + bnf i=1 n Xm X (0) (0) (n) Ψ = cj gj † + cng † , (4.18) j=1 n X X where b and c are Grassmannian c-numbers. Substituting into the action one gets
SA[Ψ, Ψ] = λncnbn , n=0 X6 so that one obtains m m (0) (0) λncnbn = [db ] [dc ] [db dc ] e− n=06 (b, c) . (4.19) hOiA i j n n O i=1 j=1 n=0 P Z Y Y Y6 This expression shows that the operator must contain exactly the number of zero- O modes of the Dirac operator that are present in the instanton background in order to have a non-vanishing vev. Hence one must study the eigenvalue problem for the Dirac operator in the instan-
ton background. The number, nL, of left-chirality zero-modes of the Dirac operator
(i.e. the number of zero-modes of iD/ L) and the number of right-chirality zero-modes
(i.e. the number of zero-modes of iD/ R), nR, are related by an index theorem [99]
n n =2drK[A] . (4.20) L − R
Moreover for configurations with self-dual (anti self-dual) field strength one has nR=0
(nL=0). Thus from (4.19) in the K-instanton sector it immediately follows that the operator must contain an excess of exactly 2Kdr factors of Ψ with respect to Ψ to O yield a non-zero result. For instance in the case of a theory containing two Weyl fermions, ψ and χ, in the fundamental representation of the gauge group SU(2) the relation
nL =2drK[A] , nR = 0 when Fµν [A]= F˜µν [A] (4.21)
gives one single zero-mode, f (0) for ψ and χ in the K=1 sector, so that
(0) (n) ψ = a0f + anf n=0 X6 (0) (n) χ = b0f + bnf , n=0 X6 4.2 Instantons and fermions 121
(n) (n) while ψ = n=0 cng † and χ = n=0 dng † do not possess zero-modes. An opera- 6 6 tor with non-vanishing vev is (x, y) = ψα(x)χ (y). The saddle point semiclassical P O P α approximation to the expectation value of is O
λn(cnan+dnbn) = [da db ] [da db dc dd ]e− n=06 hOiA 0 0 n n n n · n=0 P Z Y6 ∞ ∞ (n) (n) 2 (0) (0) a f b f = det ′(iD/ (2)) f (x)f (y) , · n n L n=0 ! n=0 ! X X where det′ denotes the product of the non-zero eigenvalues. The generalization of this elementary example to more complicated cases is rather straightforward. In the following sections computations of vev’s of composite operators in =4 supersym- N metric Yang–Mills theory will be described. The =4 theory contains four Weyl N spinors in the adjoint representation. In the case of a SU(2) gauge group equation (4.21) with d3=2 yields sixteen zero-modes in the one-instanton sector.
4.2.2 Fermionic zero-modes
The discussion of the previous subsection shows that in order to compute the in- stanton contribution to expectation values of fermionic fields one needs to know the zero-value eigenstates of the Dirac operator. In particular their norms are necessary to convert the integrations over the zero-modes into integrations over (Grassmannian) collective coordinates as in the case of the bosonic zero-modes. For a Weyl spinor in the fundamental representation of SU(2) there is one single zero mode which satisfies the equation
(f) µ αα˙ (0) iDµ (A)σ ψα =0 ,
(f) where Dµ (A) denotes the Dirac operator in the fundamental of SU(2) in the K=1 instanton sector. The solution is ε ψ(0) (x)= αs , α,s [(x x )2ρ2]3/2 − 0 where s=1,2 is an index of the fundamental of SU(2). The norm of this zero eigen- function is π ψ(0) = . k k ρ 122 Chapter 4. Non-perturbative effects in =4 SYM theory N
In the case of the adjoint representation the equation for the zero-modes becomes
iD(adj)(A) b σµ αα˙ λ(0) =0 . (4.22) µ a αb
This equation is more complicated and according to the Atiyah–Singer theorem [99] possesses four independent solutions in the case K=1 and G=SU(2). Instead of explicitly solving equation (4.22) the zero eigenfunctions can be obtained acting with the symmetries of the model 3 on the configuration
λ = λ =0 , Aµ = Aµ ,
which trivially satisfies (4.22). More precisely the zero-modes are obtained acting by the symmetries broken by the instanton background. Two of the four zero-modes are generated by a supersymmetry transformation √ρ δ λ = F σµνη 1 2 µν and are associated with the supersymmetry parameter η. The remaining two eigen- functions are obtained by a superconformal transformation with parameter ξ
1 µν κ δ2λ = F µν σ (x x0)κσ ξ . 2√ρ − The above four zero-modes can be written in the compact form 1 λ(0) = F σµν ζ (x) , (4.23) i 2 µν i where
(0) 1 µ ζi (x)= [ρηi +(x x0) σµξi 2] , i =1, 2, 3, 4 (4.24) √ρ − − and 1 0 η1 = ξ1 = , η2 = ξ2 = , η3,4 = ξ 1, 2 =0 . 0 1 − − (0) The powers of ρ in (4.24) are chosen in such a way as to give to ζi the correct dimen- 1 (0) sion, [mass]− 2 , for a supersymmetry parameter. The zero-modes λi are orthogonal and their norms are 4√2π√ρ 8π√ρ λ(0) = , j =1, 2 , λ(0) = , k =3, 4 . k j k g k k k g 3Note that this construction applies to non-supersymmetric cases as well. 4.3 Instanton calculus in SYM theories 123
The extension of the analysis described here to the case of a SU(N) gauge group, in the K=1 sector, can be obtained without too much effort and is achieved by the embedding of a SU(2) instanton in the SU(N) field configuration, see [100]. The generalization to multi-instanton sectors on the contrary is quite involved and has been developed in [101, 102, 103, 104, 105].
4.3 Instanton calculus in supersymmetric gauge theories
The machinery of the previous sections allows to compute the instanton contribution to the vacuum expectation value of composite operators in supersymmetric gauge theories. The problem is reduced to the computation of the “primed” determinants of the kinetic operators, i.e. the product of non-zero eigenvalues, and then to the integration over the collective coordinates with the measure constructed in sections 4.1 and 4.2. In non-supersymmetric theories the integration over the bosonic collective coordi- nates is IR-divergent because of the singular behaviour at ρ 0. In supersymmetric → gauge theories, on the contrary, in all the known cases the integration turns out to be finite thanks to the balance of fermionic and bosonic degrees of freedom that controls the ρ-dependence. This exact balance produces another significant simplification: the product of non-zero eigenvalues of bosonic and fermionic kinetic operators ex- actly cancels out. As a result in various models instanton calculus allows to obtain finite and exact results (in the semiclassical approximation) for the vev’s of composite operators, see [87] for a review.
Cancellation of determinants in supersymmetric theories.
Supersymmetry implies relations among the eigenvalues of the kinetic operators for scalars, vectors and spinors, that lead to an exact compensation in the ratio of de- terminants that enters the expression of the expectation values in the semiclassical approximation.
First notice that the chiral Dirac operator iD/ L(r) satisfies
2 det iD/ (r) = det iD/ (r)iD/ (r) = det iD/(r) det( 2) , L L R −